docs-samples/data-science/ai-samples/python/AIsample - Book Recommendation.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "b1cce6d9-bb3c-41be-8e30-06f2bf28020f",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "# Create, evaluate, and score a recommendation system\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d4082704-8c99-4a23-8f26-a0cde6321bed",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "## Introduction\n",
    "\n",
    "In this notebook, you walk through the Microsoft Fabric data data science workflow with an end-to-end example. The scenario is to build a book recommendation system for the users based on their preferences.\n",
    "\n",
    "A recommendation system is a class of machine learning algorithms that offers relevant suggestions to the user. Considering a user's past preferences or ratings, a recommendation system would suggest new items to the user that could include movies to watch, books to read, products to buy, etc. Several types of recommendation algorithms exist, but in this notebook you use a model-based collaborative filtering algorithm called **Alternating Least Squares (ALS) matrix factorization**.\n",
    "\n",
    "<img src=\"https://negustpublicblob.blob.core.windows.net/public/recommenders.png\" style=\"width:600px;\"/>\n",
    "\n",
    "ALS attempts to estimate a ratings matrix $R$ as a product of two lower-rank matrices known as \"factor\" matrices such that $R = U^T V$  where $U$ and $V$ can be called user and item matrix, respectively. The general approach is iterative and during each iteration, one of the factor matrices is held constant, while the other is solved using the least squares. Applying this approach alternately to the matrices $U$ and $V$ is guaranteed to monotonically decrease the overall cost function and hence you can iteratively improve the matrix factorization.\n",
    "\n",
    "<img src=\"https://negustpublicblob.blob.core.windows.net/public/Matrixfactor.svg\" style=\"width:600px;\"/>\n",
    "\n",
    "<br/><br/>\n",
    "\n",
    "Please note that this notebook can be modified to use custom data by adjusting certain parameters that will be discussed below. \n",
    "\n",
    "The summary of the main steps you take in this notebook are as following:\n",
    "\n",
    "1. Load the data\n",
    "2. Understand and process the data using exploratory data analysis\n",
    "3. Train a machine learning model using an Alternating Least Squares (ALS) model and track experiments using MLflow and Fabric Autologging feature\n",
    "4. Save the final machine learning model and load it for scoring and making predictions\n",
    "\n",
    "## Prerequisites\n",
    "- Have a lakehouse added to this notebook. You download data from a public blob, and storing that in the lakehouse. "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "62d69d41",
   "metadata": {},
   "source": [
    "## Step 1: Load the data"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cadadce4-349e-45bf-849f-7c3f61a7f734",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "### Dataset\n",
    "\n",
    "The book recommendation dataset in this scenario consists of three separate datasets:\n",
    "\n",
    "**Books.csv**\n",
    "\n",
    "Each book is identified by its International Standard Book Number (ISBN), with invalid values already being removed. Additional information (e.g., title, author, publication year, publisher) has also been obtained from Amazon Web Services. If a book has multiple authors, only the first is listed. URLs are pointing to Amazon provided for cover images in three separate formats.\n",
    "\n",
    "|ISBN|Book-Title|Book-Author|Year-Of-Publication|Publisher|Image-URL-S|Image-URL-M|Image-URL-L|\n",
    "|---|---|---|---|---|---|---|---|\n",
    "|0195153448|Classical Mythology|Mark P. O. Morford|2002|Oxford University Press|http://images.amazon.com/images/P/0195153448.01.THUMBZZZ.jpg|http://images.amazon.com/images/P/0195153448.01.MZZZZZZZ.jpg|http://images.amazon.com/images/P/0195153448.01.LZZZZZZZ.jpg|\n",
    "|0002005018|Clara Callan|Richard Bruce Wright|2001|HarperFlamingo Canada|http://images.amazon.com/images/P/0002005018.01.THUMBZZZ.jpg|http://images.amazon.com/images/P/0002005018.01.MZZZZZZZ.jpg|http://images.amazon.com/images/P/0002005018.01.LZZZZZZZ.jpg|\n",
    "\n",
    "**Ratings.csv**\n",
    "\n",
    "Ratings for each book are either explicit (provided by users, and measured on a scale from 1-10), or implicit (observed without users providing input, and indicated by 0).\n",
    "\n",
    "|User-ID|ISBN|Book-Rating|\n",
    "|---|---|---|\n",
    "|276725|034545104X|0|\n",
    "|276726|0155061224|5|\n",
    "\n",
    "**Users.csv**\n",
    "\n",
    "User IDs, which have been anonymized and mapped to integers. Demographic data (i.e. Location, Age) are provided if available or else contain *null* values.\n",
    "\n",
    "|User-ID|Location|Age|\n",
    "|---|---|---|\n",
    "|1|\"nyc| new york| usa\"||\n",
    "|2|\"stockton| california| usa\"|18.0|\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f76a2b6d",
   "metadata": {},
   "source": [
    "> [!TIP]\n",
    "> By defining the following parameters, you can use this notebook with different datasets easily."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e7b55b11-ccb1-4fbc-b689-5145002cc03c",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:33:13.0252416Z",
       "execution_start_time": "2023-09-11T14:33:11.0747126Z",
       "livy_statement_state": "available",
       "parent_msg_id": "f9e8101b-d929-4451-afda-08153df9145a",
       "queued_time": "2023-09-11T14:33:00.5636513Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": "2023-09-11T14:33:01.6760692Z",
       "spark_jobs": {
        "jobs": [],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 0,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 3
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 3, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "IS_CUSTOM_DATA = False  # if True, dataset has to be uploaded manually\n",
    "\n",
    "USER_ID_COL = \"User-ID\"  # must not be '_user_id' for this notebook to run successfully\n",
    "ITEM_ID_COL = \"ISBN\"  # must not be '_item_id' for this notebook to run successfully\n",
    "ITEM_INFO_COL = (\n",
    "    \"Book-Title\"  # must not be '_item_info' for this notebook to run successfully\n",
    ")\n",
    "RATING_COL = (\n",
    "    \"Book-Rating\"  # must not be '_rating' for this notebook to run successfully\n",
    ")\n",
    "IS_SAMPLE = True  # if True, use only <SAMPLE_ROWS> rows of data for training; otherwise use all data\n",
    "SAMPLE_ROWS = 5000  # if IS_SAMPLE is True, use only this number of rows for training\n",
    "\n",
    "DATA_FOLDER = \"Files/book-recommendation/\"  # folder containing the datasets\n",
    "ITEMS_FILE = \"Books.csv\"  # file containing the items information\n",
    "USERS_FILE = \"Users.csv\"  # file containing the users information\n",
    "RATINGS_FILE = \"Ratings.csv\"  # file containing the ratings information\n",
    "\n",
    "EXPERIMENT_NAME = \"aisample-recommendation\"  # MLflow experiment name"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ff6accde-feb6-424c-ad6d-184945dd819b",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "### Download dataset and upload to lakehouse\n",
    "\n",
    "The following code will download a publicly available version of the the dataset and then store it in a Fabric lakehouse.\n",
    "\n",
    "> [!IMPORTANT]\n",
    "> **Make sure you [add a lakehouse](https://aka.ms/fabric/addlakehouse) to the notebook before running it. Failure to do so will result in an error.**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3d7c0b6b-359e-4103-bcdc-cc568cd62bfa",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:33:13.8396711Z",
       "execution_start_time": "2023-09-11T14:33:13.5037164Z",
       "livy_statement_state": "available",
       "parent_msg_id": "1f217844-d0e1-4eb5-94bd-0d76623c5ca4",
       "queued_time": "2023-09-11T14:33:00.5675082Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 0,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 4
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 4, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Downloaded demo data files into lakehouse.\n"
     ]
    }
   ],
   "source": [
    "if not IS_CUSTOM_DATA:\n",
    "    # Download data files into lakehouse if it does not exist\n",
    "    import os, requests\n",
    "\n",
    "    remote_url = \"https://synapseaisolutionsa.blob.core.windows.net/public/Book-Recommendation-Dataset\"\n",
    "    file_list = [\"Books.csv\", \"Ratings.csv\", \"Users.csv\"]\n",
    "    download_path = f\"/lakehouse/default/{DATA_FOLDER}/raw\"\n",
    "\n",
    "    if not os.path.exists(\"/lakehouse/default\"):\n",
    "        raise FileNotFoundError(\n",
    "            \"Default lakehouse not found, please add a lakehouse and restart the session.\"\n",
    "        )\n",
    "    os.makedirs(download_path, exist_ok=True)\n",
    "    for fname in file_list:\n",
    "        if not os.path.exists(f\"{download_path}/{fname}\"):\n",
    "            r = requests.get(f\"{remote_url}/{fname}\", timeout=30)\n",
    "            with open(f\"{download_path}/{fname}\", \"wb\") as f:\n",
    "                f.write(r.content)\n",
    "    print(\"Downloaded demo data files into lakehouse.\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "abd39a7d-2eeb-41de-bde3-d5b6e866679a",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "Start recording the time it takes to run this notebook."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "1ba87a73-5ca4-4f57-9701-ab23f2fbd6e4",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:33:14.6845416Z",
       "execution_start_time": "2023-09-11T14:33:14.3283275Z",
       "livy_statement_state": "available",
       "parent_msg_id": "e7f63bd2-93a6-4d9e-9ab1-4ee92aca5ea6",
       "queued_time": "2023-09-11T14:33:00.5680881Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 0,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 5
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 5, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Record the notebook running time\n",
    "import time\n",
    "\n",
    "ts = time.time()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ca60cbe7-fb02-4f88-b9f9-c426f9bc3f7d",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "### Set up the MLflow experiment tracking\n",
    "\n",
    "Extending the MLflow autologging capabilities, autologging works by automatically capturing the values of input parameters and output metrics of a machine learning model as it is being trained. This information is then logged to your workspace, where it can be accessed and visualized using the MLflow APIs or the corresponding experiment in your workspace. For more information, see [Autologging](https://aka.ms/fabric-autologging)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "023beb35-488f-4555-aca5-7513f320e7df",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:33:18.9694569Z",
       "execution_start_time": "2023-09-11T14:33:15.2087969Z",
       "livy_statement_state": "available",
       "parent_msg_id": "b8b9e41b-24e3-4849-8dbc-7068b5284560",
       "queued_time": "2023-09-11T14:33:00.5728205Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 0,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 6
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 6, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Setup MLflow for experiment tracking\n",
    "import mlflow\n",
    "\n",
    "mlflow.set_experiment(EXPERIMENT_NAME)\n",
    "mlflow.autolog(disable=True)  # Disable MLflow autologging"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f81b3bc9-06b3-4284-ae1f-8224a6e8af15",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "> [!NOTE]\n",
    "> If you want to disable Microsoft Fabric autologging in a notebook session, call `mlflow.autolog()` and set `disable=True`."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "919804a2-0f91-42fe-a7f9-c37c6d6899c7",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "### Read raw data from the lakehouse\n",
    "\n",
    "Once the right data has landed in the Lakehouse, you can read the three separate datasets into separate Spark DataFrames in the notebook. Note that the file paths in the cell below use the parameters that you defined above."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "7afe5521-3fe1-4d15-91a7-10b18a2d7ebc",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:33:27.9671326Z",
       "execution_start_time": "2023-09-11T14:33:19.4896856Z",
       "livy_statement_state": "available",
       "parent_msg_id": "b717fe03-12d9-4429-8811-976a25a13810",
       "queued_time": "2023-09-11T14:33:00.5734525Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [
         {
          "completionTime": "2023-09-11T14:33:25.910GMT",
          "dataRead": 11148510,
          "dataWritten": 0,
          "description": "Job group for statement 7:\ndf_items = (\n    spark.read.option(\"header\", True)\n    .option(\"inferSchema\", True)\n    .csv(f\"{DATA_FOLDER}/raw/{ITEMS_FILE}\")\n    .cache()\n)\n\ndf_ratings = (\n    spark.read.option(\"header\", True)\n    .option(\"inferSchema\", True)\n    .csv(f\"{DATA_FOLDER}/raw/{RATINGS_FILE}\")\n    .cache()\n)\n\ndf_users = (\n    spark.read.option(\"header\", True)\n    .option(\"inferSchema\", True)\n    .csv(f\"{DATA_FOLDER}/raw/{USERS_FILE}\")\n    .cache()\n)",
          "displayName": "csv at NativeMethodAccessorImpl.java:0",
          "jobGroup": "7",
          "jobId": 13,
          "killedTasksSummary": {},
          "name": "csv at NativeMethodAccessorImpl.java:0",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 3,
          "numCompletedStages": 1,
          "numCompletedTasks": 3,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 3,
          "rowCount": 278860,
          "stageIds": [
           17
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:33:25.614GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:33:25.573GMT",
          "dataRead": 65536,
          "dataWritten": 0,
          "description": "Job group for statement 7:\ndf_items = (\n    spark.read.option(\"header\", True)\n    .option(\"inferSchema\", True)\n    .csv(f\"{DATA_FOLDER}/raw/{ITEMS_FILE}\")\n    .cache()\n)\n\ndf_ratings = (\n    spark.read.option(\"header\", True)\n    .option(\"inferSchema\", True)\n    .csv(f\"{DATA_FOLDER}/raw/{RATINGS_FILE}\")\n    .cache()\n)\n\ndf_users = (\n    spark.read.option(\"header\", True)\n    .option(\"inferSchema\", True)\n    .csv(f\"{DATA_FOLDER}/raw/{USERS_FILE}\")\n    .cache()\n)",
          "displayName": "csv at NativeMethodAccessorImpl.java:0",
          "jobGroup": "7",
          "jobId": 12,
          "killedTasksSummary": {},
          "name": "csv at NativeMethodAccessorImpl.java:0",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 1,
          "rowCount": 1,
          "stageIds": [
           16
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:33:25.438GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:33:25.010GMT",
          "dataRead": 22961572,
          "dataWritten": 0,
          "description": "Job group for statement 7:\ndf_items = (\n    spark.read.option(\"header\", True)\n    .option(\"inferSchema\", True)\n    .csv(f\"{DATA_FOLDER}/raw/{ITEMS_FILE}\")\n    .cache()\n)\n\ndf_ratings = (\n    spark.read.option(\"header\", True)\n    .option(\"inferSchema\", True)\n    .csv(f\"{DATA_FOLDER}/raw/{RATINGS_FILE}\")\n    .cache()\n)\n\ndf_users = (\n    spark.read.option(\"header\", True)\n    .option(\"inferSchema\", True)\n    .csv(f\"{DATA_FOLDER}/raw/{USERS_FILE}\")\n    .cache()\n)",
          "displayName": "csv at NativeMethodAccessorImpl.java:0",
          "jobGroup": "7",
          "jobId": 11,
          "killedTasksSummary": {},
          "name": "csv at NativeMethodAccessorImpl.java:0",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 6,
          "numCompletedStages": 1,
          "numCompletedTasks": 6,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 6,
          "rowCount": 1149781,
          "stageIds": [
           15
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:33:24.261GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:33:24.221GMT",
          "dataRead": 65536,
          "dataWritten": 0,
          "description": "Job group for statement 7:\ndf_items = (\n    spark.read.option(\"header\", True)\n    .option(\"inferSchema\", True)\n    .csv(f\"{DATA_FOLDER}/raw/{ITEMS_FILE}\")\n    .cache()\n)\n\ndf_ratings = (\n    spark.read.option(\"header\", True)\n    .option(\"inferSchema\", True)\n    .csv(f\"{DATA_FOLDER}/raw/{RATINGS_FILE}\")\n    .cache()\n)\n\ndf_users = (\n    spark.read.option(\"header\", True)\n    .option(\"inferSchema\", True)\n    .csv(f\"{DATA_FOLDER}/raw/{USERS_FILE}\")\n    .cache()\n)",
          "displayName": "csv at NativeMethodAccessorImpl.java:0",
          "jobGroup": "7",
          "jobId": 10,
          "killedTasksSummary": {},
          "name": "csv at NativeMethodAccessorImpl.java:0",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 1,
          "rowCount": 1,
          "stageIds": [
           14
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:33:24.045GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:33:23.570GMT",
          "dataRead": 73752387,
          "dataWritten": 0,
          "description": "Job group for statement 7:\ndf_items = (\n    spark.read.option(\"header\", True)\n    .option(\"inferSchema\", True)\n    .csv(f\"{DATA_FOLDER}/raw/{ITEMS_FILE}\")\n    .cache()\n)\n\ndf_ratings = (\n    spark.read.option(\"header\", True)\n    .option(\"inferSchema\", True)\n    .csv(f\"{DATA_FOLDER}/raw/{RATINGS_FILE}\")\n    .cache()\n)\n\ndf_users = (\n    spark.read.option(\"header\", True)\n    .option(\"inferSchema\", True)\n    .csv(f\"{DATA_FOLDER}/raw/{USERS_FILE}\")\n    .cache()\n)",
          "displayName": "csv at NativeMethodAccessorImpl.java:0",
          "jobGroup": "7",
          "jobId": 9,
          "killedTasksSummary": {},
          "name": "csv at NativeMethodAccessorImpl.java:0",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 8,
          "numCompletedStages": 1,
          "numCompletedTasks": 8,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 8,
          "rowCount": 271361,
          "stageIds": [
           13
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:33:22.215GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:33:22.083GMT",
          "dataRead": 65536,
          "dataWritten": 0,
          "description": "Job group for statement 7:\ndf_items = (\n    spark.read.option(\"header\", True)\n    .option(\"inferSchema\", True)\n    .csv(f\"{DATA_FOLDER}/raw/{ITEMS_FILE}\")\n    .cache()\n)\n\ndf_ratings = (\n    spark.read.option(\"header\", True)\n    .option(\"inferSchema\", True)\n    .csv(f\"{DATA_FOLDER}/raw/{RATINGS_FILE}\")\n    .cache()\n)\n\ndf_users = (\n    spark.read.option(\"header\", True)\n    .option(\"inferSchema\", True)\n    .csv(f\"{DATA_FOLDER}/raw/{USERS_FILE}\")\n    .cache()\n)",
          "displayName": "csv at NativeMethodAccessorImpl.java:0",
          "jobGroup": "7",
          "jobId": 8,
          "killedTasksSummary": {},
          "name": "csv at NativeMethodAccessorImpl.java:0",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 1,
          "rowCount": 1,
          "stageIds": [
           12
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:33:21.527GMT",
          "usageDescription": ""
         }
        ],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 6,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 7
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 7, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_items = (\n",
    "    spark.read.option(\"header\", True)\n",
    "    .option(\"inferSchema\", True)\n",
    "    .csv(f\"{DATA_FOLDER}/raw/{ITEMS_FILE}\")\n",
    "    .cache()\n",
    ")\n",
    "\n",
    "df_ratings = (\n",
    "    spark.read.option(\"header\", True)\n",
    "    .option(\"inferSchema\", True)\n",
    "    .csv(f\"{DATA_FOLDER}/raw/{RATINGS_FILE}\")\n",
    "    .cache()\n",
    ")\n",
    "\n",
    "df_users = (\n",
    "    spark.read.option(\"header\", True)\n",
    "    .option(\"inferSchema\", True)\n",
    "    .csv(f\"{DATA_FOLDER}/raw/{USERS_FILE}\")\n",
    "    .cache()\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d2b06290",
   "metadata": {},
   "source": [
    "## Step 2: Perform exploratory data analysis"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d1c14a18-8e5a-4a04-8412-3a564c875bc5",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "You can explore each of the DataFrames using the `display` command. This allows you to view high-level statistics of the DataFrames and understand how different columns in the datasets are related to each other. To explore the datasets, you need to import the required libraries."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "660d7422-e9f9-42f7-9d35-2d02a18c457e",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:33:32.1411961Z",
       "execution_start_time": "2023-09-11T14:33:28.4213025Z",
       "livy_statement_state": "available",
       "parent_msg_id": "3e097eca-9625-4997-a3a3-faa66b27f84a",
       "queued_time": "2023-09-11T14:33:00.5740173Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 0,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 8
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 8, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Import the required libraries\n",
    "import pyspark.sql.functions as F\n",
    "from pyspark.ml.feature import StringIndexer\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "color = sns.color_palette()  # adjusting plotting style\n",
    "import pandas as pd  # dataframes"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "30e350f6-98a4-46f3-a331-cfb063de91d7",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "### Explore the Book dataset\n",
    "Let's look into the the DataFrame that stores the book data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "f56d8d3c-c30f-4c6c-84e2-63476549ef3a",
   "metadata": {
    "advisor": {
     "adviceMetadata": "{\"artifactId\":\"974c28bc-8ec7-4ee8-8d90-830a0031857e\",\"activityId\":\"db09740a-036f-4e14-9afc-ee8d112e8535\",\"applicationId\":\"application_1694442018519_0001\",\"jobGroupId\":\"9\",\"advices\":{\"info\":1}}"
    },
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:33:39.3399249Z",
       "execution_start_time": "2023-09-11T14:33:32.7087217Z",
       "livy_statement_state": "available",
       "parent_msg_id": "427402b3-5b2f-4e94-9b8b-40aac01210b5",
       "queued_time": "2023-09-11T14:33:00.5745135Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [
         {
          "completionTime": "2023-09-11T14:33:37.040GMT",
          "dataRead": 14086341,
          "dataWritten": 0,
          "description": "Job group for statement 9:\ndisplay(df_items, summary=True)",
          "displayName": "collect at AnalysisRunner.scala:328",
          "jobGroup": "9",
          "jobId": 15,
          "killedTasksSummary": {},
          "name": "collect at AnalysisRunner.scala:328",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 1,
          "numSkippedTasks": 8,
          "numTasks": 9,
          "rowCount": 8,
          "stageIds": [
           19,
           20
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:33:36.561GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:33:36.529GMT",
          "dataRead": 73752387,
          "dataWritten": 14086341,
          "description": "Job group for statement 9:\ndisplay(df_items, summary=True)",
          "displayName": "collect at AnalysisRunner.scala:328",
          "jobGroup": "9",
          "jobId": 14,
          "killedTasksSummary": {},
          "name": "collect at AnalysisRunner.scala:328",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 8,
          "numCompletedStages": 1,
          "numCompletedTasks": 8,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 8,
          "rowCount": 271368,
          "stageIds": [
           18
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:33:33.216GMT",
          "usageDescription": ""
         }
        ],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 2,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 9
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 9, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.synapse.widget-view+json": {
       "widget_id": "67fc6032-fcee-4c3b-a40f-e4749d14ff3c",
       "widget_type": "Synapse.DataFrame"
      },
      "text/plain": [
       "SynapseWidget(Synapse.DataFrame, 67fc6032-fcee-4c3b-a40f-e4749d14ff3c)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "display(df_items, summary=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bcc1eaaa-f048-434a-90e9-048481fbf412",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "You add the column named `_item_id` for later use which must only contain integers to suit your recommendation model. Therefore, use a `StringIndexer` to transform the `ITEM_ID_COL` into indices."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "360c5b8d-ebb3-46ef-9276-f9f65d573456",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:33:42.5137889Z",
       "execution_start_time": "2023-09-11T14:33:39.8104155Z",
       "livy_statement_state": "available",
       "parent_msg_id": "877b79d3-8a41-456e-895c-2833505584d2",
       "queued_time": "2023-09-11T14:33:00.582598Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [
         {
          "completionTime": "2023-09-11T14:33:40.997GMT",
          "dataRead": 2877562,
          "dataWritten": 0,
          "description": "Job group for statement 10:\ndf_items = (\n    StringIndexer(inputCol=ITEM_ID_COL, outputCol=\"_item_id\")\n    .setHandleInvalid(\"skip\")\n    .fit(df_items)\n    .transform(df_items)\n    .withColumn(\"_item_id\", F.col(\"_item_id\").cast(\"int\"))\n)",
          "displayName": "collect at StringIndexer.scala:204",
          "jobGroup": "10",
          "jobId": 17,
          "killedTasksSummary": {},
          "name": "collect at StringIndexer.scala:204",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 1,
          "numSkippedTasks": 8,
          "numTasks": 9,
          "rowCount": 8,
          "stageIds": [
           22,
           23
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:33:40.626GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:33:40.576GMT",
          "dataRead": 74660656,
          "dataWritten": 2877562,
          "description": "Job group for statement 10:\ndf_items = (\n    StringIndexer(inputCol=ITEM_ID_COL, outputCol=\"_item_id\")\n    .setHandleInvalid(\"skip\")\n    .fit(df_items)\n    .transform(df_items)\n    .withColumn(\"_item_id\", F.col(\"_item_id\").cast(\"int\"))\n)",
          "displayName": "collect at StringIndexer.scala:204",
          "jobGroup": "10",
          "jobId": 16,
          "killedTasksSummary": {},
          "name": "collect at StringIndexer.scala:204",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 8,
          "numCompletedStages": 1,
          "numCompletedTasks": 8,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 8,
          "rowCount": 39,
          "stageIds": [
           21
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:33:39.998GMT",
          "usageDescription": ""
         }
        ],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 2,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 10
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 10, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_items = (\n",
    "    StringIndexer(inputCol=ITEM_ID_COL, outputCol=\"_item_id\")\n",
    "    .setHandleInvalid(\"skip\")\n",
    "    .fit(df_items)\n",
    "    .transform(df_items)\n",
    "    .withColumn(\"_item_id\", F.col(\"_item_id\").cast(\"int\"))\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "224e51b8-be76-437d-8964-5f7834553721",
   "metadata": {},
   "source": [
    "Take a look at these values to ensure that the content of the `_item_id` column increases successively as expected."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3234a2eb-8f97-4419-bf5a-c7ecd9bdd63f",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:33:45.7200852Z",
       "execution_start_time": "2023-09-11T14:33:43.0291028Z",
       "livy_statement_state": "available",
       "parent_msg_id": "3e0c4213-8f18-4758-bc41-a5a3c67a1355",
       "queued_time": "2023-09-11T14:33:00.5832865Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [
         {
          "completionTime": "2023-09-11T14:33:44.562GMT",
          "dataRead": 74660656,
          "dataWritten": 0,
          "description": "Job group for statement 11:\ndisplay(df_items.sort(F.col(\"_item_id\").desc()))",
          "displayName": "getRowsInJsonString at Display.scala:452",
          "jobGroup": "11",
          "jobId": 18,
          "killedTasksSummary": {},
          "name": "getRowsInJsonString at Display.scala:452",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 8,
          "numCompletedStages": 1,
          "numCompletedTasks": 8,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 8,
          "rowCount": 31,
          "stageIds": [
           24
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:33:43.362GMT",
          "usageDescription": ""
         }
        ],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 1,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 11
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 11, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.synapse.widget-view+json": {
       "widget_id": "7ee0fec7-401c-46b5-b27f-0c1d43a0d640",
       "widget_type": "Synapse.DataFrame"
      },
      "text/plain": [
       "SynapseWidget(Synapse.DataFrame, 7ee0fec7-401c-46b5-b27f-0c1d43a0d640)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "display(df_items.sort(F.col(\"_item_id\").desc()))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4655a377-f9da-4c20-b042-3b45f54e4e6a",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "##### Plot the top 10 authors with the maximum number of books"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "fd01e6ad-d2aa-4b0a-a9d5-86772453a65b",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:33:49.9422159Z",
       "execution_start_time": "2023-09-11T14:33:46.2172596Z",
       "livy_statement_state": "available",
       "parent_msg_id": "4573c0e4-5e69-4479-9274-b5d7e054fa6f",
       "queued_time": "2023-09-11T14:33:00.5985663Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [
         {
          "completionTime": "2023-09-11T14:33:47.779GMT",
          "dataRead": 74660656,
          "dataWritten": 0,
          "description": "Job group for statement 12:\ndf_books = df_items.toPandas() # Create a pandas dataframe from the spark dataframe for visualization\nplt.figure(figsize=(8,5))\nsns.countplot(y=\"Book-Author\",palette = 'Paired', data=df_books,order=df_books['Book-Author'].value_counts().index[0:10])\nplt.title(\"Top 10 authors with maximum number of books\")",
          "displayName": "toPandas at /tmp/ipykernel_9996/2332372921.py:1",
          "jobGroup": "12",
          "jobId": 19,
          "killedTasksSummary": {},
          "name": "toPandas at /tmp/ipykernel_9996/2332372921.py:1",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 8,
          "numCompletedStages": 1,
          "numCompletedTasks": 8,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 8,
          "rowCount": 31,
          "stageIds": [
           25
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:33:46.550GMT",
          "usageDescription": ""
         }
        ],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 1,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 12
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 12, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Top 10 authors with maximum number of books')"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAy4AAAHUCAYAAADcJXz9AAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjYuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/P9b71AAAACXBIWXMAAA9hAAAPYQGoP6dpAACESklEQVR4nOzdeVxO6f8/8Nfdvm9IISUttCNLQkWJyM5Y01hmjCUGYx1kG9sgjG1s2QYhyxgfJMoSUURIsiRmaphQCmk5vz/8Ol+3SovSLa/n43Ee07nOda7zPldl7nfXua4jEQRBABERERERkQyTq+wAiIiIiIiIisPEhYiIiIiIZB4TFyIiIiIiknlMXIiIiIiISOYxcSEiIiIiIpnHxIWIiIiIiGQeExciIiIiIpJ5TFyIiIiIiEjmMXEhIiIiIiKZx8SFiGSORCIp0RYWFlbhsWzbtg19+/aFpaUl5OTkYGJiUmTdjIwMjBs3DrVq1YKKigocHBywe/fuCo+xKGvWrEFgYGCB8rCwMEgkEuzbt+/zB/UZSCQS+Pv7i/u3bt2Cv78/EhMTC9R1dXWFjY3N5wuugvn7+0MikVR2GFVeYGAgJBIJoqKiKjuUErl69SpcXFygra0NiUSCgICAQuslJiZCIpHg119//bwBour/u0TlQ6GyAyAi+tCFCxek9ufOnYvTp0/j1KlTUuVWVlYVHsv27duRkpKCZs2aIS8vD9nZ2UXW7dGjBy5fvoyFCxfCwsICf/zxB/r164e8vDz079+/wmP90Jo1a1C9enX4+vp+9mtXpgsXLqBOnTri/q1btzB79my4urp+NPGsCoYNG4YOHTpUdhgkY4YMGYLMzEzs3r0burq6Vf73gKouJi5EJHNatGghtV+jRg3IyckVKP8cjh8/Djm5d4PTnTt3xo0bNwqtd/ToUYSEhIjJCgC4ubnh4cOH+Omnn/DNN99AXl7+s8UtC169egU1NbXPft3K+DmRFXXq1JFK2ujLVl6/Qzdu3MDw4cPRsWPHcoiKqPLwUTEi+iI9e/YMI0eORO3ataGkpARTU1NMnz4dWVlZUvUkEglGjx6N9evXw8LCAsrKyrCysirxI1z5SUtxDhw4AA0NDfTu3Vuq/Ntvv8U///yDyMjIj54fFRWFvn37wsTEBKqqqjAxMUG/fv3w8OFDqXpFPQqU/+hK/uNQJiYmuHnzJsLDw8VH6z78K2t2djamT5+OWrVqQUtLC+7u7oiPjy/Q9ubNm2Fvbw8VFRXo6emhe/fuiIuLk6rj6+sLDQ0NxMbGon379tDU1ES7du0AvHtMpXPnztDX14eysjJq1aqFTp064fHjx0X2x+rVqyEnJ4cnT56IZUuXLoVEIsGoUaPEsry8POjq6mLChAli2fuPigUGBorfEzc3N7EvPnyE7vLly2jdujXU1NRgamqKhQsXIi8vr8j43r/W6NGjsWXLFlhaWkJVVRWOjo64ePEiBEHAkiVLUK9ePWhoaKBt27a4e/eu1PkhISHo2rUr6tSpAxUVFZiZmeH777/Hf//9J9Z58+YNGjVqBDMzM6SlpYnlKSkpMDAwgKurK3JzcwEU/vNhYmKCzp0748iRI2jUqBFUVVXRsGFDHDlyROyjhg0bQl1dHc2aNSvw+JOrqytcXV0L3Luvr6/Uz1T+Y0ZLlizBokWLxJ9lV1dX3LlzB9nZ2ZgyZQpq1aoFbW1tdO/eXer7W5T8n627d+/Cy8sLGhoaMDIywoQJE6R+3/MfNfrwEdL8uN7/nue3efv2bXh6ekJdXR2GhoZYuHAhAODixYto1aoV1NXVYWFhga1btxYa2/Pnz/Htt99CT08P6urq8Pb2xv379wvUO3nyJNq1awctLS2oqanB2dkZoaGhUnXyv3dXrlxBr169oKuri/r163+0b27cuIGuXbtCV1dXfDz1/Vjz/13IycnB2rVrxZ//4uTl5WH+/PmoW7cuVFRU4OjoWCBeADh37hzatWsHTU1NqKmpoWXLlvjrr79KHWdR0tPT4enpiZo1a+LSpUsAgKdPn+K7776DkZERlJWVUaNGDTg7O+PkyZPFtkdfPiYuRPTFefPmDdzc3LBt2zaMHz8ef/31FwYOHIjFixejR48eBeofPnwYK1euxJw5c7Bv3z4YGxujX79+5fos9Y0bN9CwYUMoKEgPZNvZ2YnHPyYxMRGWlpYICAjA8ePHsWjRIiQnJ6Np06ZSH2JL6sCBAzA1NUWjRo1w4cIFXLhwAQcOHJCqM23aNDx8+BAbN27E77//joSEBHh7e4sfggFgwYIFGDp0KKytrREcHIwVK1bg+vXrcHJyQkJCglR7b9++RZcuXdC2bVscOnQIs2fPRmZmJjw8PPDvv/9i9erVCAkJQUBAAOrWrYuXL18WGb+7uzsEQZD6sHTy5EmoqqoiJCRELIuKisKLFy/g7u5eaDudOnXCL7/8AuBdMpTfF506dRLrpKSkYMCAARg4cCAOHz6Mjh07YurUqdixY0cJeho4cuQINm7ciIULF2LXrl14+fIlOnXqhAkTJuD8+fP47bff8Pvvv+PWrVvo2bMnBEEQz7137x6cnJywdu1anDhxAjNnzkRkZCRatWolPpaooqKCoKAgPHnyBEOGDAHw7oPlgAEDIAgCdu3aVexo3rVr1zB16lRMnjwZwcHB0NbWRo8ePTBr1ixs3LgRv/zyC3bu3Im0tDR07twZr1+/LtG9F2b16tU4f/48Vq9ejY0bN+L27dvw9vbG0KFD8fTpU2zevBmLFy/GyZMnMWzYsBK1mZ2djS5duqBdu3Y4dOgQhgwZguXLl2PRokVljjM7Oxs9evRAp06dcOjQIfH7Pm3aNAwePBhDhgzBgQMHYGlpCV9fX0RHRxdoY+jQoZCTk8Mff/yBgIAAXLp0Ca6urnjx4oVYZ8eOHWjfvj20tLSwdetWBAUFQU9PD56enoUmAz169ICZmRn27t2LdevWFRl/fHw8WrZsiZs3b2LlypUIDg6GlZUVfH19sXjxYgDvfv7zH73t1auX+PNfnN9++w3Hjh1DQEAAduzYATk5OXTs2FHq3PDwcLRt2xZpaWnYtGkTdu3aBU1NTXh7e2PPnj2lirMwjx8/RqtWrfDw4UNcuHABzZo1AwAMGjQIBw8exMyZM3HixAls3LgR7u7uSE1NLfa+qAoQiIhk3ODBgwV1dXVxf926dQIAISgoSKreokWLBADCiRMnxDIAgqqqqpCSkiKW5eTkCA0aNBDMzMxKFUenTp0EY2PjQo+Zm5sLnp6eBcr/+ecfAYDwyy+/lOpaOTk5QkZGhqCuri6sWLFCLJ81a5ZQ2D/dW7ZsEQAIDx48EMusra0FFxeXAnVPnz4tABC8vLykyoOCggQAwoULFwRBEITnz58LqqqqBeolJSUJysrKQv/+/cWywYMHCwCEzZs3S9WNiooSAAgHDx4s8b3nq1OnjjBkyBBBEAQhKytLUFdXFyZPniwAEB4+fCgIgiDMnz9fUFRUFDIyMsTzAAizZs0S9/fu3SsAEE6fPl3gGi4uLgIAITIyUqrcysqq0O/nhwAIBgYGUtc/ePCgAEBwcHAQ8vLyxPKAgAABgHD9+vVC28rLyxOys7OFhw8fCgCEQ4cOSR3fs2ePAEAICAgQZs6cKcjJyUn9rAtC4T8fxsbGgqqqqvD48WOxLCYmRgAgGBoaCpmZmQViP3z4sFjm4uJS6M/R4MGDpX4fHjx4IAAQ7O3thdzc3AL33aVLF6nzx40bJwAQ0tLSCu2P969T2O+7l5eXYGlpKe7n/1x/+H3Oj2vLli0F2ty/f79Ylp2dLdSoUUMAIFy5ckUsT01NFeTl5YXx48eLZfm/b927d5e61vnz5wUAwrx58wRBEITMzExBT09P8Pb2lqqXm5sr2NvbC82aNRPL8r93M2fO/Gh/5Ovbt6+grKwsJCUlSZV37NhRUFNTE168eCGWARBGjRpVbJv5fVWrVi3h9evXYnl6erqgp6cnuLu7i2UtWrQQ9PX1hZcvX4plOTk5go2NjVCnTh3xZ7+kceZ///bu3StcvXpVqFWrltC6dWshNTVV6jwNDQ1h3Lhxxd4LVU0ccSGiL86pU6egrq6OXr16SZXnT0L/8K+Y7dq1Q82aNcV9eXl5fPPNN7h79+5HH1cqrY89glHc4xkZGRmYPHkyzMzMoKCgAAUFBWhoaCAzM7PAY1nlpUuXLlL7+aND+Y+nXbhwAa9fvy4wud/IyAht27Yt9K/FPXv2lNo3MzODrq4uJk+ejHXr1uHWrVsljq9du3bi4x8RERF49eoVxo8fj+rVq4ujLidPnoSTkxPU1dVL3O6HDAwMxL/m5rOzsyvwmF5R3NzcpK7fsGFDAEDHjh2lvu/55e+3++TJE4wYMQJGRkZQUFCAoqIijI2NAaDA971Pnz744Ycf8NNPP2HevHmYNm0aPDw8ShSjg4MDateuXSAWV1dXqTkUhcVYWl5eXlKPWOa3+f4o1/vlSUlJxbYpkUjg7e0tVVaa71FRbXp5eYn7CgoKMDMzg6GhIRo1aiSW6+npQV9fv9BrDRgwQGq/ZcuWMDY2xunTpwG8+7l99uwZBg8ejJycHHHLy8tDhw4dcPnyZWRmZkq18eHvUFFOnTqFdu3awcjISKrc19cXr169KtHISlF69OgBFRUVcT9/JOXMmTPIzc1FZmYmIiMj0atXL2hoaIj15OXlMWjQIDx+/Fh87LS0cR4/fhytW7dGmzZtEBISAj09PanjzZo1Q2BgIObNm4eLFy9+dMEUqnqYuBDRFyc1NRUGBgYFkgF9fX0oKCgUeGTAwMCgQBv5ZeX1eEG1atUKbevZs2cAUOB/vh/q378/fvvtNwwbNgzHjx/HpUuXcPnyZdSoUeOTHtspLub3KSsrA4B4vfz7MTQ0LHBurVq1CtyvmpoatLS0pMq0tbURHh4OBwcHTJs2DdbW1qhVqxZmzZpV7AcOd3d3JCUlISEhASdPnkSjRo2gr6+Ptm3b4uTJk3j9+jUiIiKKfEyspD7sB+BdX5S03z/83iopKX20/M2bNwDePe7Vvn17BAcHY9KkSQgNDcWlS5dw8eJFACj0+kOGDEF2djYUFBTg5+dXovg+JcayqIhrqampSX2QBt59jz4lzsLaVFJSKvR3VUlJqdBrFfVvS/7vxr///gvg3WNaioqKUtuiRYsgCIL4b0S+wn7fCpOamlrk72b+8bIq6r7evn2LjIwMPH/+HIIglOj6pY3z4MGDeP36NX744Qfx36T37dmzB4MHD8bGjRvh5OQEPT09+Pj4ICUlpfQ3Sl8cripGRF+catWqITIyEoIgSCUvT548QU5ODqpXry5Vv7D/oeWXFfahtSxsbW2xa9cu5OTkSM1ziY2NBYCPviskLS0NR44cwaxZszBlyhSxPCsrq8CHmvwPWllZWVL/Uy/LPJji5PdNcnJygWP//PNPgX4ualTJ1tYWu3fvhiAIuH79OgIDAzFnzhyoqqpK3e+H8if3nzx5EiEhIeLoQrt27fDzzz/jzJkzyMrK+uTEpbLcuHED165dQ2BgIAYPHiyWfziBP19mZiYGDRoECwsL/Pvvvxg2bBgOHTpU4XGqqKhILQqQryJ+5j7F+78b76vIOIv6t8XMzAwAxN+RVatWFbna3fujwUDxo7P5qlWrVuTv5vvXLoui7ktJSQkaGhpQUFCAnJxcia5f2jiXL1+OPXv2oGPHjjhw4ADat28vdbx69eoICAhAQEAAkpKScPjwYUyZMgVPnjzBsWPHynbD9MXgiAsRfXHatWuHjIwMHDx4UKp827Zt4vH3hYaGin/5BIDc3Fzs2bMH9evXL7elY7t3746MjAzs379fqnzr1q2oVasWmjdvXuS5EokEgiAU+Ovixo0bpSbKAxBXcbp+/bpU+Z9//lmg3dKMGhTGyckJqqqqBSapP378WHz8ozQkEgns7e2xfPly6Ojo4MqVKx+tb2hoCCsrK+zfvx/R0dFi4uLh4YGnT59i2bJl0NLSQtOmTT/azocjSbIi/wPqh9/39evXF1p/xIgRSEpKQnBwMDZt2oTDhw9j+fLlFR6niYkJ7ty5I5UQpKamIiIiosKvXRpF/W4cPny4wq65c+dOqf2IiAg8fPhQXIXN2dkZOjo6uHXrFhwdHQvd8keeSqtdu3Y4deqUmADk27ZtG9TU1D5pWfDg4GCpEaaXL1/izz//ROvWrSEvLw91dXU0b94cwcHBUr9XeXl52LFjB+rUqQMLC4syxamiooLg4GB07twZXbp0+WhyXrduXYwePRoeHh7F/ntCVQNHXIjoi+Pj44PVq1dj8ODBSExMhK2tLc6dO4dffvkFXl5eBf4CX716dbRt2xYzZsyAuro61qxZg9u3b5doSeRbt26J8zJSUlLw6tUrcTUyKysr8SWYHTt2hIeHB3744Qekp6fDzMwMu3btwrFjx7Bjx46PrvqkpaWFNm3aYMmSJahevTpMTEwQHh6OTZs2QUdHR6qul5cX9PT0MHToUMyZMwcKCgoIDAzEo0ePCrSbP9KxZ88emJqaQkVFBba2tsXecz4dHR3MmDED06ZNg4+PD/r164fU1FTMnj0bKioqmDVrVrFtHDlyBGvWrEG3bt1gamoKQRAQHByMFy9elGh+Rrt27bBq1SqoqqrC2dkZAFCvXj3Uq1cPJ06cQJcuXQqs5Pah/NGu33//HZqamlBRUUG9evXKbbStrBo0aID69etjypQpEAQBenp6+PPPP6VWTcu3ceNG7NixA1u2bIG1tTWsra0xevRoTJ48Gc7OzgXm6JSnQYMGYf369Rg4cCCGDx+O1NRULF68uMBjgZXNwMAA7u7uWLBgAXR1dWFsbIzQ0FAEBwdX2DWjoqIwbNgw9O7dG48ePcL06dNRu3ZtjBw5EgCgoaGBVatWYfDgwXj27Bl69eoFfX19PH36FNeuXcPTp0+xdu3aMl171qxZOHLkCNzc3DBz5kzo6elh586d+Ouvv7B48WJoa2uX+b7k5eXh4eGB8ePHIy8vD4sWLUJ6ejpmz54t1lmwYAE8PDzg5uaGiRMnQklJCWvWrMGNGzewa9cuMTEvS5yKiorYtWsXhg0bhl69emHbtm3o168f0tLS4Obmhv79+6NBgwbQ1NTE5cuXcezYsUJXlKQqqBIXBiAiKpEPVxUThHcr/YwYMUIwNDQUFBQUBGNjY2Hq1KnCmzdvpOrh/6+ms2bNGqF+/fqCoqKi0KBBA2Hnzp0lunb+Sj+Fbe+vXCUIgvDy5UvBz89PMDAwEJSUlAQ7Ozth165dJbrO48ePhZ49ewq6urqCpqam0KFDB+HGjRuCsbGxMHjwYKm6ly5dElq2bCmoq6sLtWvXFmbNmiVs3LixwKpiiYmJQvv27QVNTU0BgLgC1Pur97yvsNWXBEEQNm7cKNjZ2QlKSkqCtra20LVrV+HmzZtSdQr7HgmCINy+fVvo16+fUL9+fUFVVVXQ1tYWmjVrJgQGBpaoXw4dOiQAEDw8PKTKhw8fLgAQVq5cWeCcwr43AQEBQr169QR5eXmpe3RxcRGsra0LtPHhillFQSGrNeX345IlS6TKC+v3W7duCR4eHoKmpqagq6sr9O7dW0hKSpK6h+vXrwuqqqoFfg7evHkjNGnSRDAxMRGeP38uCELRq4p16tTpk2LfunWr0LBhQ0FFRUWwsrIS9uzZU+SqYiW5b0H4v5W5Ll++XCC29xX1s1XYvSYnJwu9evUS9PT0BG1tbWHgwIHiynYfripWWJtF/Tx82If5sZ84cUIYNGiQoKOjI67Al5CQUOD88PBwoVOnToKenp6gqKgo1K5dW+jUqZNUn+Tfz9OnTz/aH++LjY0VvL29BW1tbUFJSUmwt7cv8PsrCKVfVWzRokXC7NmzhTp16ghKSkpCo0aNhOPHjxeof/bsWaFt27aCurq6oKqqKrRo0UL4888/yxRnYT8neXl5gp+fnyAnJyds2LBBePPmjTBixAjBzs5O0NLSElRVVQVLS0th1qxZUqvjUdUlEYT3FpQnIqpi8l9Y+Ntvv1V2KERERPQJOMeFiIiIiIhkHhMXIiIiIiKSeZycT0RVGp+GJSIiqho44kJERERERDKPiQsREREREck8Ji5ERERERCTzOMeFqqy8vDz8888/0NTUFF+ERURERESyQxAEvHz5ErVq1YKc3MfHVJi4UJX1zz//wMjIqLLDICIiIqJiPHr0CHXq1PloHSYuVGVpamoCePeLoKWlVcnREBEREdGH0tPTYWRkJH5u+xgmLlRl5T8epqWlxcSFiIiISIaV5LF+Ts4nIiIiIiKZxxEXqvL+vHoPahrFDz8SERERfe26NzGr7BCKxBEXIiIiIiKSeUxciIiIiIhI5jFxISIiIiIimcfEhYiIiIiIZB4TFyIiIiIiknlMXIiIiIiISOYxcSEiIiIiIpnHxIWIiIiIiGQeExciIiIiIpJ5TFxkTFhYGCQSCV68eFEp13d1dcW4ceNKfV55xF3WaxMRERFR1cfEpQgRERGQl5dHhw4dKuwan/uD+tu3b7F48WLY29tDTU0N1atXh7OzM7Zs2YLs7OxPartly5ZITk6GtrZ2sXWLSnKCg4Mxd+7cT4qDiIiIiKomhcoOQFZt3rwZY8aMwcaNG5GUlIS6detWdkif5O3bt/D09MS1a9cwd+5cODs7Q0tLCxcvXsSvv/6KRo0awcHBoUxtZ2dnQ0lJCQYGBp8Uo56e3iedT0RERERVF0dcCpGZmYmgoCD88MMP6Ny5MwIDAwvUOXz4MMzNzaGqqgo3Nzds3bpVahQhNTUV/fr1Q506daCmpgZbW1vs2rVLPN/X1xfh4eFYsWIFJBIJJBIJEhMTxePR0dFwdHSEmpoaWrZsifj4ePHYvXv30LVrV9SsWRMaGhpo2rQpTp48+dF7CggIwJkzZxAaGopRo0bBwcEBpqam6N+/PyIjI2Fubi7WzcvLw6RJk6CnpwcDAwP4+/tLtSWRSLBu3Tp07doV6urqmDdvXoFRlIcPH8Lb2xu6urpQV1eHtbU1jh49isTERLi5uQEAdHV1IZFI4OvrC6DgCNTbt28xadIk1K5dG+rq6mjevDnCwsI+ep9EREREVDUxcSnEnj17YGlpCUtLSwwcOBBbtmyBIAji8cTERPTq1QvdunVDTEwMvv/+e0yfPl2qjTdv3qBJkyY4cuQIbty4ge+++w6DBg1CZGQkAGDFihVwcnLC8OHDkZycjOTkZBgZGYnnT58+HUuXLkVUVBQUFBQwZMgQ8VhGRga8vLxw8uRJXL16FZ6envD29kZSUlKR97Rz5064u7ujUaNGBY4pKipCXV1d3N+6dSvU1dURGRmJxYsXY86cOQgJCZE6Z9asWejatStiY2OlYss3atQoZGVl4cyZM4iNjcWiRYugoaEBIyMj7N+/HwAQHx+P5ORkrFixotCYv/32W5w/fx67d+/G9evX0bt3b3To0AEJCQmF1s/KykJ6errURkRERERVAx8VK8SmTZswcOBAAECHDh2QkZGB0NBQuLu7AwDWrVsHS0tLLFmyBABgaWmJGzduYP78+WIbtWvXxsSJE8X9MWPG4NixY9i7dy+aN28ObW1tKCkpQU1NrdBHrObPnw8XFxcAwJQpU9CpUye8efMGKioqsLe3h729vVh33rx5OHDgAA4fPozRo0cXek8JCQlwdXUt0f3b2dlh1qxZAABzc3P89ttvCA0NhYeHh1inf//+UgnLgwcPpNpISkpCz549YWtrCwAwNTUVj+U/Eqavrw8dHZ1CY7h37x527dqFx48fo1atWgCAiRMn4tixY9iyZQt++eWXAucsWLAAs2fPLtE9EhEREdGXhSMuH4iPj8elS5fQt29fAICCggK++eYbbN68WapO06ZNpc5r1qyZ1H5ubi7mz58POzs7VKtWDRoaGjhx4sRHR0XeZ2dnJ35taGgIAHjy5AmAd4+yTZo0CVZWVtDR0YGGhgZu37790bYFQYBEIin1tfOvn3/tfI6Ojh9tw8/PD/PmzYOzszNmzZqF69evl+ja+a5cuQJBEGBhYQENDQ1xCw8Px7179wo9Z+rUqUhLSxO3R48eleqaRERERCS7OOLygU2bNiEnJwe1a9cWywRBgKKiIp4/fw5dXd1Ck4D3HyUDgKVLl2L58uUICAiAra0t1NXVMW7cOLx9+7ZEcSgqKopf518rLy8PAPDTTz/h+PHj+PXXX2FmZgZVVVX06tXro21bWFggLi6u1NfOv37+tfO9/2hZYYYNGwZPT0/89ddfOHHiBBYsWIClS5dizJgxJYohLy8P8vLyiI6Ohry8vNQxDQ2NQs9RVlaGsrJyidonIiIioi8LR1zek5OTg23btmHp0qWIiYkRt2vXrsHY2Bg7d+4EADRo0ACXL1+WOjcqKkpq/+zZs+jatSsGDhwIe3t7mJqaFpiboaSkhNzc3FLHefbsWfj6+qJ79+6wtbWFgYGB1MT+wvTv31+cE1PYfWdmZpY6juIYGRlhxIgRCA4OxoQJE7BhwwYA7+4bwEfvvVGjRsjNzcWTJ09gZmYmtX3q6mVERERE9OVh4vKeI0eO4Pnz5xg6dChsbGyktl69emHTpk0AgO+//x63b9/G5MmTcefOHQQFBYkrj+WPjpiZmSEkJAQRERGIi4vD999/j5SUFKnrmZiYIDIyEomJifjvv/8KjGoUxczMDMHBwWJS1b9//2LPHTduHJydndGuXTusXr0a165dw/379xEUFITmzZsXOeG9rMaNG4fjx4/jwYMHuHLlCk6dOoWGDRsCAIyNjSGRSHDkyBE8ffoUGRkZBc63sLDAgAED4OPjg+DgYDx48ACXL1/GokWLcPTo0XKNlYiIiIhkHxOX92zatAnu7u6FvkSxZ8+eiImJwZUrV1CvXj3s27cPwcHBsLOzw9q1a8VVxfIfVZoxYwYaN24MT09PuLq6wsDAAN26dZNqc+LEiZCXl4eVlRVq1KhR4vkvy5cvh66uLlq2bAlvb294enqicePGHz1HWVkZISEhmDRpEtavX48WLVqgadOmWLlyJfz8/GBjY1Oia5dUbm4uRo0ahYYNG6JDhw6wtLTEmjVrALxbuGD27NmYMmUKatasWeSCAlu2bIGPjw8mTJgAS0tLdOnSBZGRkVKrrxERERHR10EifDg5g8pk/vz5WLduHSeEy5D09HRoa2tjR9gVqGloVnY4RERERDKvexOzz3q9/M9raWlp0NLS+mhdTs4vozVr1qBp06aoVq0azp8/jyVLlhQ5ckBERERERJ+GiUsZJSQkYN68eXj27Bnq1q2LCRMmYOrUqZUdFhERERFRlcRHxajK4qNiRERERKUjy4+KcXI+ERERERHJPCYuREREREQk85i4EBERERGRzGPiQkREREREMo+JCxERERERyTwuh0xVnnej+sWuUkFEREREso0jLkREREREJPOYuBARERERkcxj4kJERERERDKPiQsREREREck8Ji5ERERERCTzmLgQEREREZHMY+JCREREREQyj+9xoSqvzc+7IK+sWtlhEBER0RcoeolPZYdA/x9HXIiIiIiISOYxcSEiIiIiIpnHxIWIiIiIiGQeExciIiIiIpJ5TFyIiIiIiEjmMXEhIiIiIiKZx8SFiIiIiIhkHhMXIiIiIiKSeUxciIiIiIhI5jFxISIiIiIimffFJS5hYWGQSCR48eIFACAwMBA6OjricX9/fzg4OIj7vr6+6Nat22eNsaQ+jLUsEhMTIZFIEBMTUy4xERERERHJokpLXNatWwdNTU3k5OSIZRkZGVBUVETr1q2l6p49exYSiQR37txBy5YtkZycDG1t7RJdZ8WKFQgMDCzP0Ets//79aN68ObS1taGpqQlra2tMmDChUmIhIiIiIvqSVVri4ubmhoyMDERFRYllZ8+ehYGBAS5fvoxXr16J5WFhYahVqxYsLCygpKQEAwMDSCSSEl1HW1tbakTmczl58iT69u2LXr164dKlS4iOjsb8+fPx9u3bzx7Ll4j9RERERETvq7TExdLSErVq1UJYWJhYFhYWhq5du6J+/fqIiIiQKndzcxO/fv9RseJ8+KjYsWPH0KpVK+jo6KBatWro3Lkz7t27Jx7Pf/QqKCgIrVu3hqqqKpo2bYo7d+7g8uXLcHR0hIaGBjp06ICnT58Wed0jR46gVatW+Omnn2BpaQkLCwt069YNq1atKlB3+/btMDExgba2Nvr27YuXL1+WON4P5eXlYfjw4bCwsMDDhw8BAH/++SeaNGkCFRUVmJqaYvbs2VIjXf7+/qhbty6UlZVRq1Yt+Pn5icdMTEwwd+5c9O/fHxoaGqhVq1aBe0hLS8N3330HfX19aGlpoW3btrh27Zp4/N69e+jatStq1qwJDQ0NNG3aFCdPnpRqw8TEBPPmzYOvry+0tbUxfPhwAEBERATatGkDVVVVGBkZwc/PD5mZmYXee1ZWFtLT06U2IiIiIqoaKnWOi6urK06fPi3unz59Gq6urnBxcRHL3759iwsXLoiJy6fKzMzE+PHjcfnyZYSGhkJOTg7du3dHXl6eVL1Zs2bh559/xpUrV6CgoIB+/fph0qRJWLFiBc6ePYt79+5h5syZRV7HwMAAN2/exI0bNz4az71793Dw4EEcOXIER44cQXh4OBYuXFjqeIF3fdWnTx9ERUXh3LlzMDY2xvHjxzFw4ED4+fnh1q1bWL9+PQIDAzF//nwAwL59+7B8+XKsX78eCQkJOHjwIGxtbaXaXbJkCezs7HDlyhVMnToVP/74I0JCQgAAgiCgU6dOSElJwdGjRxEdHY3GjRujXbt2ePbsGYB3jwB6eXnh5MmTuHr1Kjw9PeHt7Y2kpKQC17GxsUF0dDRmzJiB2NhYeHp6okePHrh+/Tr27NmDc+fOYfTo0YX25YIFC6CtrS1uRkZGH+17IiIiIvpyKFTmxV1dXfHjjz8iJycHr1+/xtWrV9GmTRvk5uZi5cqVAICLFy/i9evX5Za49OzZU2p/06ZN0NfXx61bt2BjYyOWT5w4EZ6engCAsWPHol+/fggNDYWzszMAYOjQoR+dOzNmzBicPXsWtra2MDY2RosWLdC+fXsMGDAAysrKYr28vDwEBgZCU1MTADBo0CCEhoaKiUVJ483IyECnTp3w+vVrhIWFiXOA5s+fjylTpmDw4MEAAFNTU8ydOxeTJk3CrFmzkJSUBAMDA7i7u0NRURF169ZFs2bNpK7p7OyMKVOmAAAsLCxw/vx5LF++HB4eHjh9+jRiY2Px5MkT8b5+/fVXHDx4EPv27cN3330He3t72Nvbi+3NmzcPBw4cwOHDh6WSkLZt22LixInivo+PD/r3749x48YBAMzNzbFy5Uq4uLhg7dq1UFFRkYpz6tSpGD9+vLifnp7O5IWIiIioiqjUERc3NzdkZmbi8uXLOHv2LCwsLKCvrw8XFxdcvnwZmZmZCAsLQ926dWFqalou17x37x769+8PU1NTaGlpoV69egBQ4K//dnZ24tc1a9YEAKmRiJo1a+LJkydFXkddXR1//fUX7t69i59//hkaGhqYMGECmjVrJjV/x8TERExaAMDQ0FCq3ZLG269fP2RkZODEiRNSCxdER0djzpw50NDQELfhw4cjOTkZr169Qu/evfH69WuYmppi+PDhOHDggNRjZADg5ORUYD8uLk5sPyMjA9WqVZO6xoMHD8RH2jIzMzFp0iRYWVlBR0cHGhoauH37doF7cHR0lNqPjo5GYGCgVLuenp7Iy8vDgwcPCvS5srIytLS0pDYiIiIiqhoqdcTFzMwMderUwenTp/H8+XO4uLgAePeYVb169XD+/HmcPn0abdu2Lbdrent7w8jICBs2bECtWrWQl5cHGxubApPBFRUVxa/zFwL4sKywx7U+VL9+fdSvXx/Dhg3D9OnTYWFhgT179uDbb78t0GZh7ZY0Xi8vL+zYsQMXL16U6q+8vDzMnj0bPXr0KBCbiooKjIyMEB8fj5CQEJw8eRIjR47EkiVLEB4eXiC2D+PMb9/Q0FBqrlK+/EURfvrpJxw/fhy//vorzMzMoKqqil69ehW4B3V1dan9vLw8fP/991JzbvLVrVu3yNiIiIiIqOqp1MQFeDfqEhYWhufPn+Onn34Sy11cXHD8+HFcvHhR/JD/qVJTUxEXF4f169eLSy6fO3euXNouCRMTE6ipqRU5ufxDpYn3hx9+gI2NDbp06YK//vpLTAIbN26M+Ph4mJmZFXkdVVVVdOnSBV26dMGoUaPQoEEDxMbGonHjxgDePa73vosXL6JBgwZi+ykpKVBQUICJiUmh7Z89exa+vr7o3r07gHePtSUmJhZ7/40bN8bNmzc/GjsRERERfR1kInEZNWoUsrOzxQ/bwLvE5YcffsCbN2/KbX6Lrq4uqlWrht9//x2GhoZISkoS526UN39/f7x69QpeXl4wNjbGixcvsHLlSmRnZ8PDw6NC4h0zZgxyc3PRuXNn/O9//0OrVq0wc+ZMdO7cGUZGRujduzfk5ORw/fp1xMbGYt68eQgMDERubi6aN28ONTU1bN++HaqqqjA2NhbbPX/+PBYvXoxu3bohJCQEe/fuxV9//QUAcHd3h5OTE7p164ZFixbB0tIS//zzD44ePYpu3brB0dERZmZmCA4Ohre3NyQSCWbMmFGi0arJkyejRYsWGDVqFIYPHw51dXXExcUhJCSk0NXZiIiIiKjqqtQ5LsC7xOX169cwMzMT55IA7xKXly9fon79+uU2wVpOTg67d+9GdHQ0bGxs8OOPP2LJkiXl0vaHXFxccP/+ffj4+KBBgwbo2LEjUlJScOLECVhaWlZYvOPGjcPs2bPh5eWFiIgIeHp64siRIwgJCUHTpk3RokULLFu2TExMdHR0sGHDBjg7O8POzg6hoaH4888/Ua1aNbHNCRMmIDo6Go0aNcLcuXOxdOlSceECiUSCo0ePok2bNhgyZAgsLCzQt29fJCYmit/P5cuXQ1dXFy1btoS3tzc8PT3F0ZyPsbOzQ3h4OBISEtC6dWs0atQIM2bMgKGhYYn6j4iIiIiqDokgCEJlB0Gyy8TEBOPGjRNX9vqSpKenQ1tbG/Zj1kFeWbWywyEiIqIvUPQSn8oOoUrL/7yWlpZW7MJKlT7iQkREREREVBwmLkREREREJPMqfXI+ybaSrP5FRERERFTROOJCREREREQyj4kLERERERHJPCYuREREREQk85i4EBERERGRzGPiQkREREREMo+rilGVd2Zev2JfaEREREREso0jLkREREREJPOYuBARERERkcxj4kJERERERDKPiQsREREREck8Ji5ERERERCTzmLgQEREREZHMY+JCREREREQyj+9xoSrvWNwGqGmoVnYYREREVIjO1iMrOwT6QnDEhYiIiIiIZB4TFyIiIiIiknlMXIiIiIiISOYxcSEiIiIiIpnHxIWIiIiIiGQeExciIiIiIpJ5TFyIiIiIiEjmMXEhIiIiIiKZx8SFiIiIiIhkHhOXKiIxMRESiQQxMTGVHUqhfH190a1bt8oOg4iIiIi+UExcytGTJ0/w/fffo27dulBWVoaBgQE8PT1x4cIFsY5EIsHBgwcrL8gKUFhSsm/fPqioqGDx4sUAgBUrViAwMPDzB0dEREREVYJCZQdQlfTs2RPZ2dnYunUrTE1N8e+//yI0NBTPnj2r7NA+q40bN2LUqFFYvXo1hg0bBgDQ1tau5KiIiIiI6EvGEZdy8uLFC5w7dw6LFi2Cm5sbjI2N0axZM0ydOhWdOnUCAJiYmAAAunfvDolEIu4DwJ9//okmTZpARUUFpqammD17NnJycsTjEokEa9euRceOHaGqqop69eph7969BeK4f/8+3NzcoKamBnt7e6nRHgCIiIhAmzZtoKqqCiMjI/j5+SEzM1M8bmJigl9++QVDhgyBpqYm6tati99//73E/bB48WKMHj0af/zxh5i0AAVHZVxdXeHn54dJkyZBT08PBgYG8Pf3l2rr9u3baNWqFVRUVGBlZYWTJ09WyRErIiIiIioeE5dyoqGhAQ0NDRw8eBBZWVmF1rl8+TIAYMuWLUhOThb3jx8/joEDB8LPzw+3bt3C+vXrERgYiPnz50udP2PGDPTs2RPXrl3DwIED0a9fP8TFxUnVmT59OiZOnIiYmBhYWFigX79+YgIUGxsLT09P9OjRA9evX8eePXtw7tw5jB49WqqNpUuXwtHREVevXsXIkSPxww8/4Pbt28X2wZQpUzB37lwcOXIEPXv2LLb+1q1boa6ujsjISCxevBhz5sxBSEgIACAvLw/dunWDmpoaIiMj8fvvv2P69OkfbS8rKwvp6elSGxERERFVDUxcyomCggICAwOxdetW6OjowNnZGdOmTcP169fFOjVq1AAA6OjowMDAQNyfP38+pkyZgsGDB8PU1BQeHh6YO3cu1q9fL3WN3r17Y9iwYbCwsMDcuXPh6OiIVatWSdWZOHEiOnXqBAsLC8yePRsPHz7E3bt3AQBLlixB//79MW7cOJibm6Nly5ZYuXIltm3bhjdv3ohteHl5YeTIkTAzM8PkyZNRvXp1hIWFffT+//e//2HRokU4dOgQ3N3dS9RndnZ2mDVrFszNzeHj4wNHR0eEhoYCAE6cOIF79+5h27ZtsLe3R6tWrQokch9asGABtLW1xc3IyKhEcRARERGR7GPiUo569uyJf/75B4cPH4anpyfCwsLQuHHjYielR0dHY86cOeKojYaGBoYPH47k5GS8evVKrOfk5CR1npOTU4ERFzs7O/FrQ0NDAO8WDci/TmBgoNR1PD09kZeXhwcPHhTahkQigYGBgdhGUezs7GBiYoKZM2fi5cuXH61b2HXy482/Tnx8PIyMjGBgYCAeb9as2Ufbmzp1KtLS0sTt0aNHJYqDiIiIiGQfJ+eXMxUVFXh4eMDDwwMzZ87EsGHDMGvWLPj6+hZ5Tl5eHmbPno0ePXoU2t7HSCQSqX1FRcUCx/Ly8sT/fv/99/Dz8yvQTt26dQttI7+d/DaKUrt2bezfvx9ubm7o0KEDjh07Bk1NzY+e87HrCIJQ4N6Ko6ysDGVl5VKdQ0RERERfBo64VDArKyupye+KiorIzc2VqtO4cWPEx8fDzMyswCYn93/foosXL0qdd/HiRTRo0KDEsTRu3Bg3b94s9DpKSkplvMP/U7duXYSHh+PJkydo3779J80xadCgAZKSkvDvv/+KZflzgoiIiIjo68PEpZykpqaibdu22LFjB65fv44HDx5g7969WLx4Mbp27SrWMzExQWhoKFJSUvD8+XMAwMyZM7Ft2zb4+/vj5s2biIuLw549e/Dzzz9LXWPv3r3YvHkz7ty5g1mzZuHSpUsFJtZ/zOTJk3HhwgWMGjUKMTExSEhIwOHDhzFmzJjy6QQAderUQVhYGFJTU9G+fXukpaWVqR0PDw/Ur18fgwcPxvXr13H+/Hlxcn5pR2KIiIiI6MvHxKWcaGhooHnz5li+fDnatGkDGxsbzJgxA8OHD8dvv/0m1lu6dClCQkJgZGSERo0aAQA8PT1x5MgRhISEoGnTpmjRogWWLVsGY2NjqWvMnj0bu3fvhp2dHbZu3YqdO3fCysqqxDHa2dkhPDwcCQkJaN26NRo1aoQZM2aIc2HKS+3atREeHo4XL17Aw8MDL168KHUb8vLyOHjwIDIyMtC0aVMMGzZMTOSKe3yOiIiIiKoeiSAIQmUHQcWTSCQ4cOBAgTfUf03Onz+PVq1a4e7du6hfv36x9dPT06GtrY09F3+FmobqZ4iQiIiISquz9cjKDoEqUf7ntbS0NGhpaX20Lifnk8w6cOAANDQ0YG5ujrt372Ls2LFwdnYuUdJCRERERFULExeSWS9fvsSkSZPw6NEjVK9eHe7u7li6dGllh0VERERElYCJyxfia3yiz8fHBz4+PpUdBhERERHJAE7OJyIiIiIimcfEhYiIiIiIZB4TFyIiIiIiknlMXIiIiIiISOYxcSEiIiIiIpnHVcWoyuvQcHixLzQiIiIiItnGERciIiIiIpJ5TFyIiIiIiEjmMXEhIiIiIiKZx8SFiIiIiIhkHhMXIiIiIiKSeUxciIiIiIhI5jFxISIiIiIimcf3uFCV12mdBxRU+aNORERFOz3mfGWHQETF4IgLERERERHJPCYuREREREQk85i4EBERERGRzGPiQkREREREMo+JCxERERERyTwmLkREREREJPOYuBARERERkcxj4kJERERERDKPiQsREREREck8Ji5ERERERCTzmLiQzHB1dcW4ceMqOwwiIiIikkFMXMpBREQE5OXl0aFDh892zbCwMEgkEujq6uLNmzdSxy5dugSJRAKJRFKqNv39/SGRSAq9j8WLF0MikcDV1fVTwgbwf7G/ePFCqjw4OBhz58795PaJiIiIqOph4lIONm/ejDFjxuDcuXNISkr6rNfW1NTEgQMHCsRTt27dMrVnaGiI06dP4/Hjx1LlW7ZsKXOb78vOzi7ymJ6eHjQ1NT/5GkRERERU9TBx+USZmZkICgrCDz/8gM6dOyMwMFDqeP7oQmhoKBwdHaGmpoaWLVsiPj5erOPv7w8HBwds374dJiYm0NbWRt++ffHy5ctirz948GBs3rxZ3H/9+jV2796NwYMHl+l+9PX10b59e2zdulUsi4iIwH///YdOnTpJ1b18+TI8PDxQvXp1aGtrw8XFBVeuXJGqI5FIsG7dOnTt2hXq6uoYNmwY3NzcAAC6urqQSCTw9fUFUPBRMRMTE/zyyy8YMmQINDU1UbduXfz+++9Fxp6VlYX09HSpjYiIiIiqBiYun2jPnj2wtLSEpaUlBg4ciC1btkAQhAL1pk+fjqVLlyIqKgoKCgoYMmSI1PF79+7h4MGDOHLkCI4cOYLw8HAsXLiw2OsPGjQIZ8+eFUd69u/fDxMTEzRu3LjM9zRkyBCpBGzz5s0YMGAAlJSUpOq9fPkSgwcPxtmzZ3Hx4kWYm5vDy8urQMI1a9YsdO3aFbGxsZgzZw72798PAIiPj0dycjJWrFhRZCxLly6Fo6Mjrl69ipEjR+KHH37A7du3C627YMECaGtri5uRkVEZe4CIiIiIZA0Tl0+0adMmDBw4EADQoUMHZGRkIDQ0tEC9+fPnw8XFBVZWVpgyZQoiIiKk5qbk5eUhMDAQNjY2aN26NQYNGlRoOx/S19dHx44dxURj8+bNBZKi0urcuTPS09Nx5swZcUSpsDbbtm2LgQMHomHDhmjYsCHWr1+PV69eITw8XKpe//79MWTIEJiamsLY2Bh6enpi7AYGBtDW1i4yFi8vL4wcORJmZmaYPHkyqlevjrCwsELrTp06FWlpaeL26NGjsncCEREREckUJi6fID4+HpcuXULfvn0BAAoKCvjmm2+kHt3KZ2dnJ35taGgIAHjy5IlYZmJiIjW/w9DQUOr4x+SPkNy/fx8XLlzAgAEDynQ/+RQVFcXRo71798LCwkIq/nxPnjzBiBEjYGFhIY5yZGRkFJjn4+joWOZY3r+uRCKBgYFBkf2irKwMLS0tqY2IiIiIqgaFyg7gS7Zp0ybk5OSgdu3aYpkgCFBUVMTz58+hq6srlisqKopf56/2lZeXV+jx/DrvH/8YLy8vfP/99xg6dCi8vb1RrVq1Mt3P+4YMGYLmzZvjxo0bRY7g+Pr64unTpwgICICxsTGUlZXh5OSEt2/fStVTV1cvcxyf0i9EREREVHVwxKWMcnJysG3bNixduhQxMTHidu3aNRgbG2Pnzp2fLRZ5eXkMGjQIYWFhn/yYWD5ra2tYW1vjxo0b6N+/f6F1zp49Cz8/P3h5ecHa2hrKysr477//im07f65Mbm5uucRKRERERFUfE5cyOnLkCJ4/f46hQ4fCxsZGauvVqxc2bdr0WeOZO3cunj59Ck9Pz0KP//3332jQoAEuXbpU4jZPnTqF5ORk6OjoFHrczMwM27dvR1xcHCIjIzFgwACoqqoW266xsTEkEgmOHDmCp0+fIiMjo8QxEREREdHXiYlLGW3atAnu7u6FTizv2bMnYmJiCiwNXJGUlJRQvXr1Il86mZ2djfj4eLx69arEbaqrqxeZtADvFgJ4/vw5GjVqhEGDBsHPzw/6+vrFtlu7dm3Mnj0bU6ZMQc2aNTF69OgSx0REREREXyeJUNjavURVQHp6OrS1tdFqUTMoqHI6FxERFe30mPOVHQLRVyn/81paWlqxCytxxIWIiIiIiGQeExciIiIiIpJ5TFyIiIiIiEjmMXEhIiIiIiKZx8SFiIiIiIhkHhMXIiIiIiKSeUxciIiIiIhI5jFxISIiIiIimce38lGV99eIkGJfaEREREREso0jLkREREREJPNKnbgIgoCHDx/i9evXFREPERERERFRAWVKXMzNzfH48eOKiIeIiIiIiKiAUicucnJyMDc3R2pqakXEQ0REREREVECZ5rgsXrwYP/30E27cuFHe8RARERERERUgEQRBKO1Jurq6ePXqFXJycqCkpARVVVWp48+ePSu3AInKKj09Hdra2khLS+OqYkREREQyqDSf18q0HHJAQEBZTiMiIiIiIiqTMiUugwcPLu84iCrMi71ByFNTq+wwiIiqPJ1+/Ss7BCKqwsr8Asrc3FwcPHgQcXFxkEgksLKyQpcuXSAvL1+e8REREREREZUtcbl79y68vLzw999/w9LSEoIg4M6dOzAyMsJff/2F+vXrl3ecRERERET0FSvTqmJ+fn6oX78+Hj16hCtXruDq1atISkpCvXr14OfnV94xEhERERHRV65MIy7h4eG4ePEi9PT0xLJq1aph4cKFcHZ2LrfgiIiIiIiIgDKOuCgrK+Ply5cFyjMyMqCkpPTJQREREREREb2vTIlL586d8d133yEyMhKCIEAQBFy8eBEjRoxAly5dyjtGIiIiIiL6ypUpcVm5ciXq168PJycnqKioQEVFBc7OzjAzM8OKFSvKO0YiIiIiIvrKlWmOi46ODg4dOoSEhATcvn0bgiDAysoKZmZm5R0fERERERFR2d/jAgDm5uYwNzcvr1iIiIiIiIgKVabEJTc3F4GBgQgNDcWTJ0+Ql5cndfzUqVPlEhxJCwwMxLhx4/DixYvKDoWIiIiI6LMq0xyXsWPHYuzYscjNzYWNjQ3s7e2ltqosJSUFY8aMgampKZSVlWFkZARvb2+EhoZWdmjlRiKR4ODBg+J+dnY2+vbtC0NDQ1y/fr3yAiMiIiKir1aZRlx2796NoKAgeHl5lXc8Mi0xMRHOzs7Q0dHB4sWLYWdnh+zsbBw/fhyjRo3C7du3y9Rubm4uJBIJ5OTKlEdWqFevXqFnz564c+cOzp07h/r161d2SERERET0FSrTJ2UlJaWvciL+yJEjIZFIcOnSJfTq1QsWFhawtrbG+PHjcfHiRbHesmXLYGtrC3V1dRgZGWHkyJHIyMgQjwcGBkJHRwdHjhyBlZUVlJWV8fDhQzx//hw+Pj7Q1dWFmpoaOnbsiISEhEJjSUxMhJycHKKioqTKV61aBWNjYwiCgLCwMEgkEoSGhsLR0RFqampo2bIl4uPjS3S/L168QPv27fH3339LJS2CIGDx4sUwNTWFqqoq7O3tsW/fPqlzb926BS8vL2hoaKBmzZoYNGgQ/vvvP/G4q6sr/Pz8MGnSJOjp6cHAwAD+/v5SbaSlpeG7776Dvr4+tLS00LZtW1y7dq1EsRMRERFR1VKmxGXChAlYsWIFBEEo73hk1rNnz3Ds2DGMGjUK6urqBY7r6OiIX8vJyWHlypW4ceMGtm7dilOnTmHSpElS9V+9eoUFCxZg48aNuHnzJvT19eHr64uoqCgcPnwYFy5cgCAI8PLyQnZ2doHrmZiYwN3dHVu2bJEq37JlC3x9fSGRSMSy6dOnY+nSpYiKioKCggKGDBlS7P2mpKTAxcUFeXl5CA8Ph6GhoXjs559/xpYtW7B27VrcvHkTP/74IwYOHIjw8HAAQHJyMlxcXODg4ICoqCgcO3YM//77L/r06SN1ja1bt0JdXR2RkZFYvHgx5syZg5CQEADvkqNOnTohJSUFR48eRXR0NBo3box27drh2bNnhcaclZWF9PR0qY2IiIiIqgaJUMLso0ePHlL7p06dgp6eHqytraGoqCh1LDg4uPwilBGXLl1C8+bNERwcjO7du5fq3L179+KHH34QRxwCAwPx7bffIiYmRpwTlJCQAAsLC5w/fx4tW7YEAKSmpsLIyAhbt25F7969C0zODwoKwogRI5CcnAxlZWVcu3YNjRo1wv3792FiYoKwsDC4ubnh5MmTaNeuHQDg6NGj6NSpE16/fg0VFZVC45VIJFBSUoKpqSmio6OhpqYmHsvMzET16tVx6tQpODk5ieXDhg3Dq1ev8Mcff2DmzJmIjIzE8ePHxeOPHz+GkZER4uPjYWFhAVdXV+Tm5uLs2bNinWbNmqFt27ZYuHAhTp06he7du+PJkydQVlYW65iZmWHSpEn47rvvCsTt7++P2bNnFyh/uHEDtN67ByIiqhg6/fpXdghE9IVJT0+HtrY20tLSoKWl9dG6JR5x0dbWltq6d+8OFxcXVK9evcCxqig/v3t/JKMop0+fhoeHB2rXrg1NTU34+PggNTUVmZmZYh0lJSXY2dmJ+3FxcVBQUEDz5s3FsmrVqsHS0hJxcXGFXqdbt25QUFDAgQMHAACbN2+Gm5sbTExMpOq9f538kZMnT5589B68vb1x584drF+/Xqr81q1bePPmDTw8PKChoSFu27Ztw7179wAA0dHROH36tNTxBg0aAIBY58O48mPLjys6OhoZGRmoVq2aVDsPHjyQauN9U6dORVpamrg9evToo/dIRERERF+OEk/O//CRpK+Nubk5JBIJ4uLi0K1btyLrPXz4EF5eXhgxYgTmzp0LPT09nDt3DkOHDpV65EtVVVUqCSpq4EsQhCKTJSUlJQwaNAhbtmxBjx498McffyAgIKBAvfdHxPLb+nAJ6w8NHDgQXbp0wZAhQ5Cbm4uJEydKnffXX3+hdu3aUufkj4zk5eXB29sbixYtKtDu+4+cfThSJ5FIxPbz8vJgaGiIsLCwAm28/1jeh9d/f3SGiIiIiKqOMq0q1rZtWwQHBxf4AJmeno5u3bpVyfe46OnpwdPTE6tXr4afn1+BeS4vXryAjo4OoqKikJOTg6VLl4qrhAUFBRXbvpWVFXJychAZGSn1qNidO3fQsGHDIs8bNmwYbGxssGbNGmRnZxd4pO9T+Pj4QF5eHoMHD0ZeXh4mTZokLiaQlJQEFxeXQs9r3Lgx9u/fDxMTEygolO0dp40bN0ZKSgoUFBQKjCARERER0denTJPzw8LC8Pbt2wLlb968kZqzUNWsWbMGubm5aNasGfbv34+EhATExcVh5cqV4nyP+vXrIycnB6tWrcL9+/exfft2rFu3rti2zc3N0bVrVwwfPhznzp3DtWvXMHDgQNSuXRtdu3Yt8ryGDRuiRYsWmDx5Mvr16wdVVdVyu18AGDBgALZv345p06Zh4cKF0NTUxMSJE/Hjjz9i69atuHfvHq5evYrVq1dj69atAIBRo0bh2bNn6NevHy5duoT79+/jxIkT4uhNSbi7u8PJyQndunXD8ePHkZiYiIiICPz8888FVlIjIiIioqqvVH8Of//lg7du3UJKSoq4n5ubi2PHjhV4fKgqqVevHq5cuYL58+djwoQJSE5ORo0aNdCkSROsXbsWAODg4IBly5Zh0aJFmDp1Ktq0aYMFCxbAx8en2Pa3bNmCsWPHonPnznj79i3atGmDo0ePFnik6kNDhw5FREREiVYLK4t+/fpBXl4eAwYMQF5eHubOnQt9fX0sWLAA9+/fh46ODho3boxp06YBAGrVqoXz589j8uTJ8PT0RFZWFoyNjdGhQ4cSv6tGIpHg6NGjmD59OoYMGYKnT5/CwMAAbdq0Qc2aNSvkPomIiIhIdpV4VTHg3TK/+XMkCjtNVVUVq1atqrAP0FS4+fPnY/fu3YiNja3sUGRK/ioVXFWMiOjz4KpiRFRapVlVrFQjLg8ePIAgCDA1NcWlS5dQo0YN8ZiSkhL09fUhLy9ftqip1DIyMhAXF4dVq1Zh7ty5lR0OEREREVGFKVXiYmxsDKD4Fano8xg9ejR27dqFbt26cZSLiIiIiKq0Mi35tG3bto8eL8l8Dvp0gYGBCAwMrOwwiIiIiIgqXJkSl7Fjx0rtZ2dn49WrV1BSUoKamhoTFyIiIiIiKldlWg75+fPnUltGRgbi4+PRqlUr7Nq1q7xjJCIiIiKir1yZEpfCmJubY+HChQVGY4iIiIiIiD5VuSUuACAvL49//vmnPJskIiIiIiIq2xyXw4cPS+0LgoDk5GT89ttvcHZ2LpfAiIiIiIiI8pXqBZT5Pnz7uUQiQY0aNdC2bVssXboUhoaG5RYgUVmV5oVGRERERPT5VdgLKPPxPS5ERERERPQ5lescl9jYWIwbN648myQiIiIiIvr0xCU9PR3r169Hs2bNYG9vj7CwsHIIi4iIiIiI6P+UOXEJDw+Hj48PDA0NMXLkSLRt2xZ37txBTExMOYZHRERERERUysQlOTkZv/zyC8zMzNC3b19Ur14d4eHhkJOTg4+PD8zMzCoqTiIiIiIi+oqVanJ+vXr10Lt3b6xevRoeHh4FVhcjIiIiIiKqCKXKPIyNjXHu3DmcOXMGd+7cqaiYiIiIiIiIpJRqxCU+Ph7nz5/Hpk2b0LRpU1hYWGDgwIEA3r3LhUgWRbu4QUNevrLDICKq0ppGXarsEIioiiv1s17Ozs7YvHkzkpOTMWLECAQFBSE3NxcjR47Ehg0b8PTp04qIk4iIiIiIvmJlnqSioaGB4cOH48KFC7h58yaaNGmCn3/+GbVq1SrP+IiIiIiIiMrnBZQNGzbEr7/+ir///ht79uwpjyaJiIiIiIhEn5y4aGlp4f79+wAABQUF9OjR45ODIiIiIiIiet8nJy6CIJRHHEREREREREXii1iIiIiIiEjmfXLiMnDgQGhpaZVHLERERERERIUqU+Ly6NEj8eu1a9eievXq4v7Fixc/PSoiIiIiIqL3lClx8fDwQGpqaoHy8+fPo0OHDp8cFBERERER0fvKlLi0bt0a7du3x8uXL8WyM2fOwMvLC7NmzSq34IiIiIiIiIAyJi6///476tWrh06dOuHNmzc4ffo0OnXqhDlz5uDHH38s7xgJgEQiwcGDBys7jErl6+uLbt26VXYYRERERFQJypS4SCQS7Nq1CyoqKmjXrh26dOmCBQsWYOzYseUd32fn6+sLiURSYLt7926lxpWcnIyOHTtW6DUCAwOl7tnQ0BB9+vTBgwcPKvS6RERERETFUShpxevXrxcomzVrFvr164eBAweiTZs2Yh07O7vyi7ASdOjQAVu2bJEqq1GjRoF6b9++hZKS0meJycDA4LNcR0tLC/Hx8RAEAbdv38b333+PLl26ICYmBvLy8p8lBiIiIiKiD5V4xMXBwQGNGjWCg4ODuLVp0waPHz/G+vXrxWONGjWqyHg/C2VlZRgYGEht8vLycHV1xejRozF+/HhUr14dHh4eAIBly5bB1tYW6urqMDIywsiRI5GRkSG2FxgYCB0dHRw/fhwNGzaEhoYGOnTogOTkZKnrbt68GdbW1lBWVoahoSFGjx4tHnv/UbHExERIJBIEBwfDzc0NampqsLe3x4ULF6Tai4iIQJs2baCqqgojIyP4+fkhMzPzo/cukUhgYGAAQ0NDuLm5YdasWbhx4wbu3r2Ly5cvw8PDA9WrV4e2tjZcXFxw5coVqfP9/f1Rt25dKCsro1atWvDz8xOPZWVlYdKkSTAyMoKysjLMzc2xadMmAEBubi6GDh2KevXqQVVVFZaWllixYkUJv2P/1356errURkRERERVQ4kTlwcPHuD+/ft48OCBuL2/n//1/fv3KzLeSrd161YoKCjg/PnzWL9+PQBATk4OK1euxI0bN7B161acOnUKkyZNkjrv1atX+PXXX7F9+3acOXMGSUlJmDhxonh87dq1GDVqFL777jvExsbi8OHDMDMz+2gs06dPx8SJExETEwMLCwv069cPOTk5AIDY2Fh4enqiR48euH79Ovbs2YNz585JJUMloaqqCgDIzs7Gy5cvMXjwYJw9exYXL16Eubk5vLy8xEUa9u3bh+XLl2P9+vVISEjAwYMHYWtrK7bl4+OD3bt3Y+XKlYiLi8O6deugoaEBAMjLy0OdOnUQFBSEW7duYebMmZg2bRqCgoJKHOuCBQugra0tbkZGRqW6VyIiIiKSXRJBEITKDkKW+Pr6YseOHVBRURHLOnbsiL1798LV1RVpaWm4evXqR9vYu3cvfvjhB/z3338A3o24fPvtt7h79y7q168PAFizZg3mzJmDlJQUAEDt2rXx7bffYt68eYW2KZFIcODAAXTr1g2JiYmoV68eNm7ciKFDhwIAbt26BWtra8TFxaFBgwbw8fGBqqqqmFwBwLlz5+Di4oLMzEyp+8sXGBiIcePG4cWLFwCAx48fo3fv3nj8+DHu3btX4LG43Nxc6Orq4o8//kDnzp2xbNkyrF+/Hjdu3ICioqJU3Tt37sDS0hIhISFwd3f/aP/lGzVqFP7991/s27cPwLvvzYsXL4pcpCArKwtZWVnifnp6OoyMjHDKoTE0+JgbEVGFahp1qbJDIKIvUHp6OrS1tZGWllbsS+1LPMflQ/fu3UNAQADi4uIgkUjQsGFDjB07Vvxg/iVzc3PD2rVrxX11dXXxa0dHxwL1T58+jV9++QW3bt1Ceno6cnJy8ObNG2RmZornqqmpSfWNoaEhnjx5AgB48uQJ/vnnH7Rr165Ucb4/l8jQ0FBsq0GDBoiOjsbdu3exc+dOsY4gCMjLy8ODBw/QsGHDQttMS0uDhoYGBEHAq1ev0LhxYwQHB0NJSQlPnjzBzJkzcerUKfz777/Izc3Fq1evkJSUBADo3bs3AgICYGpqig4dOsDLywve3t5QUFAQ58i4uLgUeT/r1q3Dxo0b8fDhQ7x+/Rpv376Fg4NDiftDWVkZysrKJa5PRERERF+OMiUux48fR5cuXeDg4ABnZ2cIgoCIiAhYW1vjzz//FOd+fKnU1dWLfEzr/SQGAB4+fAgvLy+MGDECc+fOhZ6eHs6dO4ehQ4ciOztbrPfhCIREIkH+YFf+41il9X6bEokEwLtHrvL/+/3330vNMclXt27dItvU1NTElStXICcnh5o1a0rdr6+vL54+fYqAgAAYGxtDWVkZTk5OePv2LQDAyMgI8fHxCAkJwcmTJzFy5EgsWbIE4eHhxd5jUFAQfvzxRyxduhROTk7Q1NTEkiVLEBkZWfIOISIiIqIqq0yJy5QpU/Djjz9i4cKFBconT578xScupREVFYWcnBwsXboUcnLvpgyVZl4G8C5ZMDExQWhoKNzc3MolrsaNG+PmzZvFzpP5kJycXJHnnD17FmvWrIGXlxcA4NGjR+LjcPlUVVXRpUsXdOnSBaNGjUKDBg0QGxsLW1tb5OXlITw8vNBHxc6ePYuWLVti5MiRYtm9e/dKFTsRERERVV1leo9LXFycOLfifUOGDMGtW7c+OagvSf369ZGTk4NVq1bh/v372L59O9atW1fqdvz9/bF06VKsXLkSCQkJuHLlClatWlXmuCZPnowLFy5g1KhRiImJQUJCAg4fPowxY8aUuU0zMzNs374dcXFxiIyMxIABA6RGUgIDA7Fp0ybcuHFD7AtVVVUYGxvDxMQEgwcPxpAhQ3Dw4EE8ePAAYWFhYpJnZmaGqKgoHD9+HHfu3MGMGTNw+fLlMsdKRERERFVLmRKXGjVqICYmpkB5TEwM9PX1PzWmL4qDgwOWLVuGRYsWwcbGBjt37sSCBQtK3c7gwYMREBCANWvWwNraGp07d0ZCQkKZ47Kzs0N4eDgSEhLQunVrNGrUCDNmzBDnwpTF5s2b8fz5czRq1AiDBg2Cn5+f1PdbR0cHGzZsgLOzM+zs7BAaGoo///wT1apVA/Bu5bRevXph5MiRaNCgAYYPHy4uzzxixAj06NED33zzDZo3b47U1FSp0RciIiIi+rqVaVWxOXPmYPny5ZgyZQpatmwJiUSCc+fOYdGiRZgwYQJ+/vnnioiVqFTyV6ngqmJERBWPq4oRUVlU+KpiM2bMgKamJpYuXYqpU6cCAGrVqgV/f/9CJ4MTERERERF9ik9+j0v+ywc1NTXLJSCi8sIRFyKiz4cjLkRUFp/lPS4A8PTpU8THx0MikcDS0hLVq1f/lOaIiIiIiIgKVabJ+ZmZmRgyZAgMDQ3Rpk0btG7dGoaGhhg6dChevXpV3jESEREREdFXrkyJy/jx4xEeHo4///wTL168wIsXL3Do0CGEh4djwoQJ5R0jERERERF95cr0qNj+/fuxb98+uLq6imVeXl5QVVVFnz59sHbt2vKKj4iIiIiIqGwjLq9evULNmjULlOvr6/NRMSIiIiIiKndlSlycnJwwa9YsvHnzRix7/fo1Zs+eDScnp3ILjoiIiIiICCjjcsg3btxAhw4d8ObNG9jb20MikSAmJgbKyso4ceIErK2tKyJWolIpzfJ6RERERPT5lebzWpnf4/L69Wvs2LEDt2/fhiAIsLKywoABA6CqqlqmoInKGxMXIiIiItlWms9rZXpULDU1Faqqqhg+fDjGjh0LDQ0NxMfHIyoqqkwBExERERERfUypEpfY2FiYmJhAX18fDRo0QExMDJo1a4bly5fj999/h5ubGw4ePFhBoRIRERER0deqVInLpEmTYGtri/DwcLi6uqJz587w8vJCWloanj9/ju+//x4LFy6sqFiJiIiIiOgrVao5LtWrV8epU6dgZ2eHjIwMaGlp4dKlS3B0dAQA3L59Gy1atMCLFy8qKl6iEuMcFyIiIiLZVmFzXJ49ewYDAwMAgIaGBtTV1aGnpyce19XVxcuXL8sQMhERERERUdFKPTlfIpF8dJ+IiIiIiKi8KZT2BF9fXygrKwMA3rx5gxEjRkBdXR0AkJWVVb7REZWD1PCFeKuuUtlhEBF9suptZ1Z2CERElaZUicvgwYOl9gcOHFigjo+Pz6dFRERERERE9IFSJS5btmypqDiIiIiIiIiKVKYXUBIREREREX1OTFyIiIiIiEjmMXEhIiIiIiKZx8SFiIiIiIhkHhMXIiIiIiKSeUxciIiIiIhI5jFxISIiIiIimcfEhUrN1dUV48aNq+wwiIiIiOgrUqoXUFLxfH198eLFCxw8eLCyQymWpaUlHjx4gAcPHqB27dolPi84OBiKiooVGBkRERERkTSOuHylzp07hzdv3qB3794IDAws1bl6enrQ1NSsmMCIiIiIiArBxKWC7du3D7a2tlBVVUW1atXg7u6OzMxMAMDly5fh4eGB6tWrQ1tbGy4uLrhy5YrU+cuWLYOtrS3U1dVhZGSEkSNHIiMjQ6rO+fPn4eLiAjU1Nejq6sLT0xPPnz//aFybNm1C//79MWjQIGzevBmCIEgdX7NmDczNzaGiooKaNWuiV69e4rEPHxUzMTHBvHnz4OPjAw0NDRgbG+PQoUN4+vQpunbtCg0NDdja2iIqKkrqGvv374e1tTWUlZVhYmKCpUuXisemTp2KFi1aFIjbzs4Os2bN+ui9EREREVHVw8SlAiUnJ6Nfv34YMmQI4uLiEBYWhh49eohJwsuXLzF48GCcPXsWFy9ehLm5Oby8vPDy5UuxDTk5OaxcuRI3btzA1q1bcerUKUyaNEk8HhMTg3bt2sHa2hoXLlzAuXPn4O3tjdzc3CLjevnyJfbu3YuBAwfCw8MDmZmZCAsLE49HRUXBz88Pc+bMQXx8PI4dO4Y2bdp89F6XL18OZ2dnXL16FZ06dcKgQYPg4+ODgQMH4sqVKzAzM4OPj49479HR0ejTpw/69u2L2NhY+Pv7Y8aMGeLoz4ABAxAZGYl79+6J17h58yZiY2MxYMCAQmPIyspCenq61EZEREREVYNE+PBP7fRJ3p/jcuXKFTRp0gSJiYkwNjYu9tzc3Fzo6urijz/+QOfOnQuts3fvXvzwww/477//AAD9+/dHUlISzp07V+IYN2zYgDVr1uDq1asAgHHjxuG///7Djh07ALybw/Ltt9/i8ePHhT4S5urqCgcHBwQEBAB4N+LSunVrbN++HQCQkpICQ0NDzJgxA3PmzAEAXLx4EU5OTkhOToaBgQEGDBiAp0+f4sSJE2K7kyZNwl9//YWbN28CAOzt7dGrVy/MmDEDADBt2jScPHkSly5dKvS+/P39MXv27ALl9w9Phaa6Son7h4hIVlVvO7OyQyAiKlfp6enQ1tZGWloatLS0PlqXIy4VyN7eHu3atYOtrS169+6NDRs2SD3C9eTJE4wYMQIWFhbQ1taGtrY2MjIykJSUJNY5ffo0PDw8ULt2bWhqasLHxwepqani42b5Iy6lsWnTJgwcOFDcHzhwIIKDg/HixQsAgIeHB4yNjWFqaopBgwZh586dePXq1UfbtLOzE7+uWbMmAMDW1rZA2ZMnTwAAcXFxcHZ2lmrD2dkZCQkJ4mjRgAEDsHPnTgCAIAjYtWtXkaMtwLvHy9LS0sTt0aNHH42ZiIiIiL4cTFwqkLy8PEJCQvC///0PVlZWWLVqlbiSF/BudCY6OhoBAQGIiIhATEwMqlWrhrdv3wIAHj58CC8vL9jY2GD//v2Ijo7G6tWrAQDZ2dkAAFVV1VLFdOvWLURGRmLSpElQUFCAgoICWrRogdevX2PXrl0AAE1NTVy5cgW7du2CoaEhZs6cCXt7ezGxKcz7q4xJJJIiy/Ly8gC8S0Tyy/J9OPjXv39/3LlzB1euXEFERAQePXqEvn37FhmDsrIytLS0pDYiIiIiqhqYuFQwiUQCZ2dnzJ49G1evXoWSkhIOHDgAADh79iz8/Pzg5eUlTlLPfwQMeDfXJCcnB0uXLkWLFi1gYWGBf/75R6p9Ozs7hIaGljieTZs2oU2bNrh27RpiYmLEbdKkSdi0aZNYT0FBAe7u7li8eDGuX7+OxMREnDp16hN74/9YWVkVeLwtIiICFhYWkJeXBwDUqVMHbdq0wc6dO7Fz5064u7uLIzdERERE9HXhe1wqUGRkJEJDQ9G+fXvo6+sjMjIST58+RcOGDQEAZmZm2L59OxwdHZGeno6ffvpJagSlfv36yMnJwapVq+Dt7Y3z589j3bp1UteYOnUqbG1tMXLkSIwYMQJKSko4ffo0evfujerVq0vVzc7Oxvbt2zFnzhzY2NhIHRs2bBgWL16Ma9eu4dGjR7h//z7atGkDXV1dHD16FHl5ebC0tCy3vpkwYQKaNm2KuXPn4ptvvsGFCxfw22+/Yc2aNVL1BgwYAH9/f7x9+xbLly8vt+sTERER0ZeFIy4VSEtLC2fOnIGXlxcsLCzw888/Y+nSpejYsSMAYPPmzXj+/DkaNWqEQYMGwc/PD/r6+uL5Dg4OWLZsGRYtWgQbGxvs3LkTCxYskLqGhYUFTpw4gWvXrqFZs2ZwcnLCoUOHoKBQMCc9fPgwUlNT0b179wLHzM3NYWtri02bNkFHRwfBwcFo27YtGjZsiHXr1mHXrl2wtrYut75p3LgxgoKCsHv3btjY2GDmzJmYM2cOfH19per17t0bqampePXqFbp161Zu1yciIiKiLwtXFaMqK3+VCq4qRkRVBVcVI6KqhquKERERERFRlcLEhYiIiIiIZB4TFyIiIiIiknlMXIiIiIiISOYxcSEiIiIiIpnHxIWIiIiIiGQeExciIiIiIpJ5TFyIiIiIiEjmFXy9OlEVU81lSrEvNCIiIiIi2cYRFyIiIiIiknlMXIiIiIiISOYxcSEiIiIiIpnHxIWIiIiIiGQeExciIiIiIpJ5TFyIiIiIiEjmMXEhIiIiIiKZx/e4UJX3cIEtNJWZoxOR7DDxf1DZIRARfXH4aY6IiIiIiGQeExciIiIiIpJ5TFyIiIiIiEjmMXEhIiIiIiKZx8SFiIiIiIhkHhMXIiIiIiKSeUxciIiIiIhI5jFxISIiIiIimcfEhYiIiIiIZB4TFyIiIiIiknlMXKjE/P394eDgUNlhEBEREdFXiIlLOfP19YVEIsHChQulyg8ePAiJRFJJUQESiUTcNDQ0YG9vj8DAwEqL50OBgYHQ0dGp7DCIiIiISEYxcakAKioqWLRoEZ4/f16u7b59+/aTzt+yZQuSk5Nx7do1fPPNN/j2229x/Pjxcoqu7LKzsys7BCIiIiKScUxcKoC7uzsMDAywYMGCj9bbv38/rK2toaysDBMTEyxdulTquImJCebNmwdfX19oa2tj+PDhAIDJkyfDwsICampqMDU1xYwZM0r04V9HRwcGBgaoX78+pk2bBj09PZw4cUI8npSUhK5du0JDQwNaWlro06cP/v333wLtrF+/HkZGRlBTU0Pv3r3x4sULqeNbtmxBw4YNoaKiggYNGmDNmjXiscTEREgkEgQFBcHV1RUqKirYsWMHvv32W6SlpYmjQv7+/gCANWvWwNzcHCoqKqhZsyZ69epV5P1lZWUhPT1daiMiIiKiqoGJSwWQl5fHL7/8glWrVuHx48eF1omOjkafPn3Qt29fxMbGwt/fHzNmzCjw+NaSJUtgY2OD6OhozJgxAwCgqamJwMBA3Lp1CytWrMCGDRuwfPnyEseXm5uLoKAgPHv2DIqKigAAQRDQrVs3PHv2DOHh4QgJCcG9e/fwzTffSJ179+5dBAUF4c8//8SxY8cQExODUaNGicc3bNiA6dOnY/78+YiLi8Mvv/yCGTNmYOvWrVLtTJ48GX5+foiLi0O7du0QEBAALS0tJCcnIzk5GRMnTkRUVBT8/PwwZ84cxMfH49ixY2jTpk2R97VgwQJoa2uLm5GRUYn7hIiIiIhkm0QQBKGyg6hKfH198eLFCxw8eBBOTk6wsrLCpk2bcPDgQXTv3h353T1gwAA8ffpUasRj0qRJ+Ouvv3Dz5k0A70ZcGjVqhAMHDnz0mkuWLMGePXsQFRVVZB2JRAIVFRXIy8vjzZs3yM3NhZ6eHiIjI2FmZoaQkBB07NgRDx48ED/w37p1C9bW1rh06RKaNm0Kf39/zJs3D4mJiahTpw4A4NixY+jUqRP+/vtvGBgYoG7duli0aBH69esnXnvevHk4evQoIiIikJiYiHr16iEgIABjx44V6wQGBmLcuHFSozfBwcH49ttv8fjxY2hqahbb91lZWcjKyhL309PTYWRkhOtT6kJTmTk6EckOE/8HlR0CEZFMSE9Ph7a2NtLS0qClpfXRuvw0V4EWLVqErVu34tatWwWOxcXFwdnZWarM2dkZCQkJyM3NFcscHR0LnLtv3z60atUKBgYG0NDQwIwZM5CUlFRsPMuXL0dMTAxCQkLg4OCA5cuXw8zMTIzHyMhIapTCysoKOjo6iIuLE8vq1q0rJi0A4OTkhLy8PMTHx+Pp06d49OgRhg4dCg0NDXGbN28e7t27JxVLYff1IQ8PDxgbG8PU1BSDBg3Czp078erVqyLrKysrQ0tLS2ojIiIioqqBiUsFatOmDTw9PTFt2rQCxwRBKLDKWGGDX+rq6lL7Fy9eRN++fdGxY0ccOXIEV69exfTp00s0cd/AwABmZmZwc3PD3r17MWrUKDGpKiyej5Xnyz8mkUiQl5cH4N3jYjExMeJ248YNXLx48aP3VRhNTU1cuXIFu3btgqGhIWbOnAl7e/sCc2qIiIiIqOpj4lLBFi5ciD///BMRERFS5VZWVjh37pxUWUREBCwsLCAvL19ke+fPn4exsTGmT58OR0dHmJub4+HDh6WOy8zMDD179sTUqVPFeJKSkvDo0SOxzq1bt5CWloaGDRuKZUlJSfjnn3/E/QsXLkBOTg4WFhaoWbMmateujfv378PMzExqq1ev3kfjUVJSkhppyqegoAB3d3csXrwY169fR2JiIk6dOlXq+yUiIiKiL5tCZQdQ1dna2mLAgAFYtWqVVPmECRPQtGlTzJ07F9988w0uXLiA3377TWoFrsKYmZkhKSkJu3fvRtOmTfHXX38VOwemKBMmTIC9vT2ioqLg7u4OOzs7DBgwAAEBAcjJycHIkSPh4uIi9ViXiooKBg8ejF9//RXp6enw8/NDnz59YGBgAODdSyr9/PygpaWFjh07IisrC1FRUXj+/DnGjx9fZCwmJibIyMhAaGgo7O3toaamhlOnTuH+/fto06YNdHV1cfToUeTl5cHS0rJM90tEREREXy6OuHwGc+fOLfAYWOPGjREUFITdu3fDxsYGM2fOxJw5c+Dr6/vRtrp27Yoff/wRo0ePhoODAyIiIsTVxkrL1tYW7u7umDlzJiQSCQ4ePAhdXV20adMG7u7uMDU1xZ49e6TOMTMzQ48ePeDl5YX27dvDxsZGKtkaNmwYNm7ciMDAQNja2sLFxQWBgYHFjri0bNkSI0aMwDfffIMaNWpg8eLF0NHRQXBwMNq2bYuGDRti3bp12LVrF6ytrct0v0RERET05eKqYlRl5a9SwVXFiEjWcFUxIqJ3uKoYERERERFVKUxciIiIiIhI5jFxISIiIiIimcfEhYiIiIiIZB4TFyIiIiIiknlMXIiIiIiISOYxcSEiIiIiIpnHxIWIiIiIiGSeQmUHQFTRjKfGFvtCIyIiIiKSbRxxISIiIiIimcfEhYiIiIiIZB4TFyIiIiIiknlMXIiIiIiISOYxcSEiIiIiIpnHxIWIiIiIiGQeExciIiIiIpJ5fI8LVXlRITehrqZR2WEQ0VekeUfbyg6BiKjK4YgLERERERHJPCYuREREREQk85i4EBERERGRzGPiQkREREREMo+JCxERERERyTwmLkREREREJPOYuBARERERkcxj4kJERERERDKPiQsREREREck8Ji6fwNXVFePGjavsMCqdiYkJAgICKvw6vr6+6NatW4Vfh4iIiIhkT5VOXHx9fSGRSMStWrVq6NChA65fv17ZoVWYt2/fYvHixbC3t4eamhqqV68OZ2dnbNmyBdnZ2Z/UdmBgIHR0dMonUCIiIiKiUqjSiQsAdOjQAcnJyUhOTkZoaCgUFBTQuXPnT2rzUxOAjxEEATk5OWU69+3bt/D09MTChQvx3XffISIiApcuXcKoUaOwatUq3Lx5s8xxVeQ9ExEREREVp8onLsrKyjAwMICBgQEcHBwwefJkPHr0CE+fPhXrTJ48GRYWFlBTU4OpqSlmzJgh9UHd398fDg4O2Lx5M0xNTaGsrAxBEAAAOTk5GD16NHR0dFCtWjX8/PPP4jEA2LFjBxwdHaGpqQkDAwP0798fT548EY+HhYVBIpHg+PHjcHR0hLKyMs6ePYt79+6ha9euqFmzJjQ0NNC0aVOcPHnyo/caEBCAM2fOIDQ0FKNGjYKDgwNMTU3Rv39/REZGwtzcHABw7NgxtGrVSoy5c+fOuHfvnthOYmIiJBIJgoKC4OrqChUVFezYsQPffvst0tLSxBEsf3//QuNYtmwZbG1toa6uDiMjI4wcORIZGRni8fyRm+PHj6Nhw4bQ0NAQE8x8ubm5GD9+vBjjpEmTpPqViIiIiL4uVT5xeV9GRgZ27twJMzMzVKtWTSzX1NREYGAgbt26hRUrVmDDhg1Yvny51Ll3795FUFAQ9u/fj5iYGLF869atUFBQQGRkJFauXInly5dj48aN4vG3b99i7ty5uHbtGg4ePIgHDx7A19e3QGyTJk3CggULEBcXBzs7O2RkZMDLywsnT57E1atX4enpCW9vbyQlJRV5fzt37oS7uzsaNWpU4JiioiLU1dUBAJmZmRg/fjwuX76M0NBQyMnJoXv37sjLy5M6Z/LkyfDz80NcXBzatWuHgIAAaGlpiSNYEydOLDQOOTk5rFy5Ejdu3MDWrVtx6tQpTJo0SarOq1ev8Ouvv2L79u04c+YMkpKSpNpbunQpNm/ejE2bNuHcuXN49uwZDhw4UOS9A0BWVhbS09OlNiIiIiKqGhQqO4CKduTIEWhoaAB494Hd0NAQR44cgZzc/+VsP//8s/i1iYkJJkyYgD179kh92H779i22b9+OGjVqSLVvZGSE5cuXQyKRwNLSErGxsVi+fDmGDx8OABgyZIhY19TUFCtXrkSzZs2QkZEhxgUAc+bMgYeHh7hfrVo12Nvbi/vz5s3DgQMHcPjwYYwePbrQe01ISICrq2uxfdKzZ0+p/U2bNkFfXx+3bt2CjY2NWD5u3Dj06NFD3NfW1oZEIoGBgcFH239/wYJ69eph7ty5+OGHH7BmzRqxPDs7G+vWrUP9+vUBAKNHj8acOXPE4wEBAZg6daoY67p163D8+PGPXnfBggWYPXv2R+sQERER0Zepyo+4uLm5ISYmBjExMYiMjET79u3RsWNHPHz4UKyzb98+tGrVCgYGBtDQ0MCMGTMKjGwYGxsXSFoAoEWLFpBIJOK+k5MTEhISkJubCwC4evUqunbtCmNjY2hqaoqJxYftOzo6Su1nZmZi0qRJsLKygo6ODjQ0NHD79u2PjrgIgiAVS1Hu3buH/v37w9TUFFpaWqhXr16JYiqp06dPw8PDA7Vr14ampiZ8fHyQmpqKzMxMsY6ampqYtACAoaGh+AhdWloakpOT4eTkJB5XUFAoNp6pU6ciLS1N3B49elSm+ImIiIhI9lT5xEVdXR1mZmYwMzNDs2bNsGnTJmRmZmLDhg0AgIsXL6Jv377o2LEjjhw5gqtXr2L69Ol4+/ZtgXZKKzMzE+3bt4eGhgZ27NiBy5cvi487Fdf+Tz/9hP3792P+/Pk4e/YsYmJiYGtrW+C891lYWCAuLq7YuLy9vZGamooNGzYgMjISkZGRJYqpJB4+fAgvLy/Y2Nhg//79iI6OxurVqwFIT/BXVFSUOk8ikXzyHBZlZWVoaWlJbURERERUNVT5R8U+JJFIICcnh9evXwMAzp8/D2NjY0yfPl2s8/5oTHEuXrxYYN/c3Bzy8vK4ffs2/vvvPyxcuBBGRkYAgKioqBK1e/bsWfj6+qJ79+4A3s3PSUxM/Og5/fv3x7Rp03D16tUC81xycnKQlZWFN2/eIC4uDuvXr0fr1q0BAOfOnStRTEpKSuJIUlGioqKQk5ODpUuXio/jBQUFlaj9fNra2jA0NMTFixfRpk0bMf7o6Gg0bty4VG0RERERUdVQ5UdcsrKykJKSgpSUFMTFxWHMmDHIyMiAt7c3AMDMzAxJSUnYvXs37t27h5UrVxY7Cfx9jx49wvjx4xEfH49du3Zh1apVGDt2LACgbt26UFJSwqpVq3D//n0cPnwYc+fOLVG7ZmZmCA4ORkxMDK5du4b+/fsXmDz/oXHjxsHZ2Rnt2rXD6tWrce3aNdy/fx9BQUFo3rw5EhISoKuri2rVquH333/H3bt3cerUKYwfP75EMZmYmCAjIwOhoaH477//8OrVqwJ16tevj5ycHPGet2/fjnXr1pWo/feNHTsWCxcuxIEDB3D79m2MHDkSL168KHU7RERERFQ1VPnE5dixYzA0NIShoSGaN2+Oy5cvY+/eveJck65du+LHH3/E6NGj4eDggIiICMyYMaPE7fv4+OD169do1qwZRo0ahTFjxuC7774DANSoUQOBgYHYu3cvrKyssHDhQvz6668lanf58uXQ1dVFy5Yt4e3tDU9Pz2JHG5SVlRESEoJJkyZh/fr1aNGiBZo2bYqVK1fCz88PNjY2kJOTw+7duxEdHQ0bGxv8+OOPWLJkSYliatmyJUaMGIFvvvkGNWrUwOLFiwvUcXBwwLJly7Bo0SLY2Nhg586dWLBgQYnaf9+ECRPg4+MDX19fODk5QVNTUxx9IiIiIqKvj0TgyzGoikpPT4e2tjZC90VAXU2j+BOIiMpJ8462lR0CEdEXIf/zWlpaWrHzk6v8iAsREREREX35mLgQEREREZHMY+JCREREREQyj4kLERERERHJPCYuREREREQk85i4EBERERGRzGPiQkREREREMo+JCxERERERyTyFyg6AqKI5elgX+0IjIiIiIpJtHHEhIiIiIiKZx8SFiIiIiIhkHhMXIiIiIiKSeUxciIiIiIhI5jFxISIiIiIimcfEhYiIiIiIZB4TFyIiIiIiknl8jwtVeRN6roaSokplh0FEVdjqoz9WdghERFUeR1yIiIiIiEjmMXEhIiIiIiKZx8SFiIiIiIhkHhMXIiIiIiKSeUxciIiIiIhI5jFxISIiIiIimcfEhYiIiIiIZB4TFyIiIiIiknlMXIiIiIiISOYxcSEiIiIiIpnHxKWEJBIJDh48WCFtm5iYICAgoELaLo2wsDBIJBK8ePGiRPVdXV0xbty4Co2JiIiIiAhg4gIASElJwZgxY2BqagplZWUYGRnB29sboaGhlR1auTAxMYFEIoFEIoGqqipMTEzQp08fnDp1Sqpey5YtkZycDG1t7UqKlIiIiIiocF994pKYmIgmTZrg1KlTWLx4MWJjY3Hs2DG4ublh1KhRFXrtt2/fVmj775szZw6Sk5MRHx+Pbdu2QUdHB+7u7pg/f75YR0lJCQYGBpBIJJ8tLiIiIiKikvjqE5eRI0dCIpHg0qVL6NWrFywsLGBtbY3x48fj4sWLUnX/++8/dO/eHWpqajA3N8fhw4fFY7m5uRg6dCjq1asHVVVVWFpaYsWKFVLn+/r6olu3bliwYAFq1aoFCwuLQmNKS0vDd999B319fWhpaaFt27a4du2aePzatWtwc3ODpqYmtLS00KRJE0RFRX30PjU1NWFgYIC6deuiTZs2+P333zFjxgzMnDkT8fHxAAp/VOz8+fNwcXGBmpoadHV14enpiefPnxd6jWPHjkFbWxvbtm0DAPz999/45ptvoKuri2rVqqFr165ITEws0B+//vorDA0NUa1aNYwaNQrZ2dlinTVr1sDc3BwqKiqoWbMmevXqVeQ9ZmVlIT09XWojIiIioqrhq05cnj17hmPHjmHUqFFQV1cvcFxHR0dqf/bs2ejTpw+uX78OLy8vDBgwAM+ePQMA5OXloU6dOggKCsKtW7cwc+ZMTJs2DUFBQVJthIaGIi4uDiEhIThy5EiBawqCgE6dOiElJQVHjx5FdHQ0GjdujHbt2onXGjBgAOrUqYPLly8jOjoaU6ZMgaKiYqnvf+zYsRAEAYcOHSr0eExMDNq1awdra2tcuHAB586dg7e3N3JzcwvU3b17N/r06YNt27bBx8cHr169gpubGzQ0NHDmzBmcO3cOGhoa6NChg9RI0+nTp3Hv3j2cPn0aW7duRWBgIAIDAwEAUVFR8PPzw5w5cxAfH49jx46hTZs2Rd7PggULoK2tLW5GRkal7hMiIiIikk0KlR1AZbp79y4EQUCDBg1KVN/X1xf9+vUDAPzyyy9YtWoVLl26hA4dOkBRURGzZ88W69arVw8REREICgpCnz59xHJ1dXVs3LgRSkpKhV7j9OnTiI2NxZMnT6CsrAwA+PXXX3Hw4EHs27cP3333HZKSkvDTTz+JcZubm5fp/vX09KCvry81CvK+xYsXw9HREWvWrBHLrK2tC9Rbs2YNpk2bhkOHDsHNzQ3Au0RGTk4OGzduFB8927JlC3R0dBAWFob27dsDAHR1dfHbb79BXl4eDRo0QKdOnRAaGorhw4cjKSkJ6urq6Ny5MzQ1NWFsbIxGjRoVeT9Tp07F+PHjxf309HQmL0RERERVxFeduAiCAAAlntNhZ2cnfq2urg5NTU08efJELFu3bh02btyIhw8f4vXr13j79i0cHByk2rC1tS0yaQGA6OhoZGRkoFq1alLlr1+/xr179wAA48ePx7Bhw7B9+3a4u7ujd+/eqF+/fonu4UOCIBR5/zExMejdu/dHz9+/fz/+/fdfnDt3Ds2aNZO6j7t370JTU1Oq/ps3b8T7AN4lQvLy8uK+oaEhYmNjAQAeHh4wNjaGqakpOnTogA4dOoiP6hVGWVlZTPaIiIiIqGr5qh8VMzc3h0QiQVxcXInqf/g4lkQiQV5eHgAgKCgIP/74I4YMGYITJ04gJiYG3377bYEJ+IU9kva+vLw8GBoaIiYmRmqLj4/HTz/9BADw9/fHzZs30alTJ5w6dQpWVlY4cOBASW9blJqaiqdPn6JevXqFHldVVS22DQcHB9SoUQNbtmwRE8H8+2jSpEmB+7hz5w769+8v1vtYn2pqauLKlSvYtWsXDA0NMXPmTNjb25d4uWYiIiIiqjq+6sRFT08Pnp6eWL16NTIzMwscL80H5LNnz6Jly5YYOXIkGjVqBDMzM6mRhZJq3LgxUlJSoKCgADMzM6mtevXqYj0LCwv8+OOPOHHiBHr06IEtW7aU+lorVqyAnJwcunXrVuhxOzu7YpeErl+/Pk6fPo1Dhw5hzJgxUveRkJAAfX39AvdRmuWWFRQU4O7ujsWLF+P69etITEwssIwzEREREVV9X3XiArybn5Gbm4tmzZph//79SEhIQFxcHFauXAknJ6cSt2NmZoaoqCgcP34cd+7cwYwZM3D58uVSx+Pu7g4nJyd069YNx48fR2JiIiIiIvDzzz8jKioKr1+/xujRoxEWFoaHDx/i/PnzuHz5Mho2bPjRdl++fImUlBQ8evQIZ86cwXfffYd58+Zh/vz5MDMzK/ScqVOn4vLlyxg5ciSuX7+O27dvY+3atfjvv/+k6llYWOD06dPYv3+/+ELKAQMGoHr16ujatSvOnj2LBw8eIDw8HGPHjsXjx49L1BdHjhzBypUrERMTg4cPH2Lbtm3Iy8uDpaVlic4nIiIioqrjq09c6tWrhytXrsDNzQ0TJkyAjY0NPDw8EBoairVr15a4nREjRqBHjx745ptv0Lx5c6SmpmLkyJGljkcikeDo0aNo06YNhgwZAgsLC/Tt2xeJiYmoWbMm5OXlkZqaCh8fH1hYWKBPnz7o2LGj1MIAhZk5cyYMDQ1hZmaGQYMGIS0tDaGhoZg8eXKR51hYWODEiRO4du0amjVrBicnJxw6dAgKCgWnRllaWuLUqVPYtWsXJkyYADU1NZw5cwZ169ZFjx490LBhQwwZMgSvX7+GlpZWifpCR0cHwcHBaNu2LRo2bIh169Zh165dhS4QQERERERVm0R4f2ICURWSnp4ObW1tDHP/BUqKKpUdDhFVYauP/ljZIRARfZHyP6+lpaUV+8ftr37EhYiIiIiIZB8TFyIiIiIiknlMXIiIiIiISOYxcSEiIiIiIpnHxIWIiIiIiGQeExciIiIiIpJ5TFyIiIiIiEjmMXEhIiIiIiKZV/AV6ERVzNL9o4p9oRERERERyTaOuBARERERkcxj4kJERERERDKPj4pRlSUIAgAgPT29kiMhIiIiosLkf07L/9z2MUxcqMpKTU0FABgZGVVyJERERET0MS9fvoS2tvZH6zBxoSpLT08PAJCUlFTsLwKVn/T0dBgZGeHRo0dcFOEzYZ9XDvb758c+rxzs98rxtfS7IAh4+fIlatWqVWxdJi5UZcnJvZvCpa2tXaV/4WWVlpYW+/0zY59XDvb758c+rxzs98rxNfR7Sf/AzMn5REREREQk85i4EBERERGRzGPiQlWWsrIyZs2aBWVl5coO5avCfv/82OeVg/3++bHPKwf7vXKw3wuSCCVZe4yIiIiIiKgSccSFiIiIiIhkHhMXIiIiIiKSeUxciIiIiIhI5jFxISIiIiIimcfEhaqsNWvWoF69elBRUUGTJk1w9uzZyg7pi3XmzBl4e3ujVq1akEgkOHjwoNRxQRDg7++PWrVqQVVVFa6urrh586ZUnaysLIwZMwbVq1eHuro6unTpgsePH3/Gu/iyLFiwAE2bNoWmpib09fXRrVs3xMfHS9Vhv5e/tWvXws7OTnzhm5OTE/73v/+Jx9nnFW/BggWQSCQYN26cWMZ+L3/+/v6QSCRSm4GBgXicfV4x/v77bwwcOBDVqlWDmpoaHBwcEB0dLR5nv38cExeqkvbs2YNx48Zh+vTpuHr1Klq3bo2OHTsiKSmpskP7ImVmZsLe3h6//fZboccXL16MZcuW4bfffsPly5dhYGAADw8PvHz5Uqwzbtw4HDhwALt378a5c+eQkZGBzp07Izc393PdxhclPDwco0aNwsWLFxESEoKcnBy0b98emZmZYh32e/mrU6cOFi5ciKioKERFRaFt27bo2rWr+MGBfV6xLl++jN9//x12dnZS5ez3imFtbY3k5GRxi42NFY+xz8vf8+fP4ezsDEVFRfzvf//DrVu3sHTpUujo6Ih12O/FEIiqoGbNmgkjRoyQKmvQoIEwZcqUSoqo6gAgHDhwQNzPy8sTDAwMhIULF4plb968EbS1tYV169YJgiAIL168EBQVFYXdu3eLdf7++29BTk5OOHbs2GeL/Uv25MkTAYAQHh4uCAL7/XPS1dUVNm7cyD6vYC9fvhTMzc2FkJAQwcXFRRg7dqwgCPxZryizZs0S7O3tCz3GPq8YkydPFlq1alXkcfZ78TjiQlXO27dvER0djfbt20uVt2/fHhEREZUUVdX14MEDpKSkSPW3srIyXFxcxP6Ojo5Gdna2VJ1atWrBxsaG35MSSktLAwDo6ekBYL9/Drm5udi9ezcyMzPh5OTEPq9go0aNQqdOneDu7i5Vzn6vOAkJCahVqxbq1auHvn374v79+wDY5xXl8OHDcHR0RO/evaGvr49GjRphw4YN4nH2e/GYuFCV899//yE3Nxc1a9aUKq9ZsyZSUlIqKaqqK79PP9bfKSkpUFJSgq6ubpF1qGiCIGD8+PFo1aoVbGxsALDfK1JsbCw0NDSgrKyMESNG4MCBA7CysmKfV6Ddu3fjypUrWLBgQYFj7PeK0bx5c2zbtg3Hjx/Hhg0bkJKSgpYtWyI1NZV9XkHu37+PtWvXwtzcHMePH8eIESPg5+eHbdu2AeDPekkoVHYARBVFIpFI7QuCUKCMyk9Z+pvfk5IZPXo0rl+/jnPnzhU4xn4vf5aWloiJicGLFy+wf/9+DB48GOHh4eJx9nn5evToEcaOHYsTJ05ARUWlyHrs9/LVsWNH8WtbW1s4OTmhfv362Lp1K1q0aAGAfV7e8vLy4OjoiF9++QUA0KhRI9y8eRNr166Fj4+PWI/9XjSOuFCVU716dcjLyxf4y8OTJ08K/BWDPl3+KjQf628DAwO8ffsWz58/L7IOFW7MmDE4fPgwTp8+jTp16ojl7PeKo6SkBDMzMzg6OmLBggWwt7fHihUr2OcVJDo6Gk+ePEGTJk2goKAABQUFhIeHY+XKlVBQUBD7jf1esdTV1WFra4uEhAT+rFcQQ0NDWFlZSZU1bNhQXDiI/V48Ji5U5SgpKaFJkyYICQmRKg8JCUHLli0rKaqqq169ejAwMJDq77dv3yI8PFzs7yZNmkBRUVGqTnJyMm7cuMHvSREEQcDo0aMRHByMU6dOoV69elLH2e+fjyAIyMrKYp9XkHbt2iE2NhYxMTHi5ujoiAEDBiAmJgampqbs988gKysLcXFxMDQ05M96BXF2di6wrP2dO3dgbGwMgP+ul8jnXw+AqOLt3r1bUFRUFDZt2iTcunVLGDdunKCuri4kJiZWdmhfpJcvXwpXr14Vrl69KgAQli1bJly9elV4+PChIAiCsHDhQkFbW1sIDg4WYmNjhX79+gmGhoZCenq62MaIESOEOnXqCCdPnhSuXLkitG3bVrC3txdycnIq67Zk2g//r537CYmij+M4/plldR1DomRz9WBJf7CIoqhTYeZeIowiAg9RG3sQXAwCITGwTgZ7iCjWg0YEUl2CLiFUGC5BIHZI0sqI/tBlyUMUlRLRfp/DwzM8Zk+0saNTz/sFA7vznd/O7/c97PJhdqatzRYvXmzZbNZyuZy3TU9Pe8fQ9+Lr6uqyu3fv2suXL+3hw4d24sQJC4VCdvv2bTOj5/Pl308VM6Pvfujo6LBsNmsvXrywkZERa25utoqKCu93kp4X3+joqIXDYevp6bFnz57ZlStXrLy83C5fvuwdQ99/jOCCP1Zvb68tX77cSktLbfPmzd5jZFG44eFhkzRnSyQSZvb3IxxPnTplsVjMIpGINTQ02Pj4+KzPmJmZsfb2dlu6dKm5rmvNzc32+vXrBVjN7+F7/ZZkly5d8o6h78WXTCa9741oNGrxeNwLLWb0fL58G1zoe/G1tLRYdXW1lZSUWE1Nje3fv98ePXrk1em5P27cuGHr16+3SCRi9fX11t/fP6tO33/MMTNbmGs9AAAAAPBzuMcFAAAAQOARXAAAAAAEHsEFAAAAQOARXAAAAAAEHsEFAAAAQOARXAAAAAAEHsEFAAAAQOARXAAAAAAEHsEFAAAAQOARXAAA+B949eqVHMfR2NjYQk8FAH4JwQUAAABA4BFcAACYB/l8Xul0WqtWrVIkElFtba16enokSePj42pqapLruqqsrFRra6s+fvzojW1sbNSxY8dmfd6+fft05MgR7/2KFSt0+vRpJZNJVVRUqLa2Vv39/V69rq5OkrRp0yY5jqPGxkbf1goAfiC4AAAwD7q6upROp9Xd3a3Hjx/r6tWrqqqq0vT0tHbt2qUlS5bo/v37unbtmoaGhtTe3l7wOc6cOaMtW7bowYMHSqVSamtr0+TkpCRpdHRUkjQ0NKRcLqfr168XdX0A4LfwQk8AAIA/3YcPH3Tu3DllMhklEglJ0sqVK7V9+3ZduHBBMzMzGhgY0KJFiyRJmUxGe/bsUTqdVlVV1U+fZ/fu3UqlUpKkzs5OnT17VtlsVvX19YpGo5KkyspKxWKxIq8QAPzHFRcAAHz25MkTff78WfF4/Lu1jRs3eqFFkrZt26Z8Pq+nT58WdJ4NGzZ4rx3HUSwW09TU1K9PHAAChOACAIDPXNf9z5qZyXGc79b+2R8KhWRms2pfvnyZc3xJScmc8fl8vtDpAkAgEVwAAPDZ6tWr5bqu7ty5M6e2bt06jY2N6dOnT96+e/fuKRQKac2aNZKkaDSqXC7n1b9+/aqJiYmC5lBaWuqNBYDfEcEFAACflZWVqbOzU8ePH9fAwICeP3+ukZERXbx4UQcPHlRZWZkSiYQmJiY0PDyso0eP6tChQ979LU1NTRocHNTg4KAmJyeVSqX07t27guawbNkyua6rmzdv6s2bN3r//r0PKwUA/xBcAACYB93d3ero6NDJkye1du1atbS0aGpqSuXl5bp165bevn2rrVu36sCBA4rH48pkMt7YZDKpRCKhw4cPa8eOHaqrq9POnTsLOn84HNb58+fV19enmpoa7d27t9hLBABfOfbtn2YBAAAAIGC44gIAAAAg8AguAAAAAAKP4AIAAAAg8AguAAAAAAKP4AIAAAAg8AguAAAAAAKP4AIAAAAg8AguAAAAAAKP4AIAAAAg8AguAAAAAAKP4AIAAAAg8P4CJVlf/xBCYuMAAAAASUVORK5CYII=",
      "text/plain": [
       "<Figure size 800x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_books = df_items.toPandas() # Create a pandas dataframe from the spark dataframe for visualization\n",
    "plt.figure(figsize=(8,5))\n",
    "sns.countplot(y=\"Book-Author\",palette = 'Paired', data=df_books,order=df_books['Book-Author'].value_counts().index[0:10])\n",
    "plt.title(\"Top 10 authors with maximum number of books\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cf7a1f6e-b930-4adf-a952-d821a9aeb5d9",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "Agatha Christie is the leading author with over 600 books, followed by William Shakespeare."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "71577c6c-1e24-4556-bc5e-20757849b529",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "### Explore the User dataset\n",
    "Take a look at the dataframe that stores the user data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3493804b-6faa-46ab-8f00-f27ef985a09e",
   "metadata": {
    "advisor": {
     "adviceMetadata": "{\"artifactId\":\"974c28bc-8ec7-4ee8-8d90-830a0031857e\",\"activityId\":\"db09740a-036f-4e14-9afc-ee8d112e8535\",\"applicationId\":\"application_1694442018519_0001\",\"jobGroupId\":\"13\",\"advices\":{\"info\":1}}"
    },
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:33:53.0188199Z",
       "execution_start_time": "2023-09-11T14:33:50.4043725Z",
       "livy_statement_state": "available",
       "parent_msg_id": "233d4e50-df57-44db-9796-ae1ac0afa12b",
       "queued_time": "2023-09-11T14:33:00.5992608Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [
         {
          "completionTime": "2023-09-11T14:33:52.023GMT",
          "dataRead": 1572878,
          "dataWritten": 0,
          "description": "Job group for statement 13:\ndisplay(df_users, summary=True)",
          "displayName": "collect at AnalysisRunner.scala:328",
          "jobGroup": "13",
          "jobId": 21,
          "killedTasksSummary": {},
          "name": "collect at AnalysisRunner.scala:328",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 1,
          "numSkippedTasks": 3,
          "numTasks": 4,
          "rowCount": 3,
          "stageIds": [
           27,
           28
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:33:51.883GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:33:51.859GMT",
          "dataRead": 11148510,
          "dataWritten": 1572878,
          "description": "Job group for statement 13:\ndisplay(df_users, summary=True)",
          "displayName": "collect at AnalysisRunner.scala:328",
          "jobGroup": "13",
          "jobId": 20,
          "killedTasksSummary": {},
          "name": "collect at AnalysisRunner.scala:328",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 3,
          "numCompletedStages": 1,
          "numCompletedTasks": 3,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 3,
          "rowCount": 278862,
          "stageIds": [
           26
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:33:50.730GMT",
          "usageDescription": ""
         }
        ],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 2,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 13
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 13, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.synapse.widget-view+json": {
       "widget_id": "3b55c788-a0d5-411f-b7d0-14677f8a281f",
       "widget_type": "Synapse.DataFrame"
      },
      "text/plain": [
       "SynapseWidget(Synapse.DataFrame, 3b55c788-a0d5-411f-b7d0-14677f8a281f)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "display(df_users, summary=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "829fc998-b2f0-4761-819f-717ad967d7f5",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "Remove the rows with missing `User-ID` values to clean up the data prior to training."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3bcfb117-d761-467a-95d1-445937f3e598",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:33:55.2342996Z",
       "execution_start_time": "2023-09-11T14:33:53.5723606Z",
       "livy_statement_state": "available",
       "parent_msg_id": "f85aaec5-1ec7-445f-8905-e391575003b5",
       "queued_time": "2023-09-11T14:33:00.5999422Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [
         {
          "completionTime": "2023-09-11T14:33:54.497GMT",
          "dataRead": 1572875,
          "dataWritten": 0,
          "description": "Job group for statement 14:\ndf_users = df_users.dropna(subset=(USER_ID_COL))\ndisplay(df_users, summary=True)",
          "displayName": "collect at AnalysisRunner.scala:328",
          "jobGroup": "14",
          "jobId": 23,
          "killedTasksSummary": {},
          "name": "collect at AnalysisRunner.scala:328",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 1,
          "numSkippedTasks": 3,
          "numTasks": 4,
          "rowCount": 3,
          "stageIds": [
           30,
           31
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:33:54.415GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:33:54.396GMT",
          "dataRead": 6319048,
          "dataWritten": 1572875,
          "description": "Job group for statement 14:\ndf_users = df_users.dropna(subset=(USER_ID_COL))\ndisplay(df_users, summary=True)",
          "displayName": "collect at AnalysisRunner.scala:328",
          "jobGroup": "14",
          "jobId": 22,
          "killedTasksSummary": {},
          "name": "collect at AnalysisRunner.scala:328",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 3,
          "numCompletedStages": 1,
          "numCompletedTasks": 3,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 3,
          "rowCount": 32,
          "stageIds": [
           29
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:33:53.817GMT",
          "usageDescription": ""
         }
        ],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 2,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 14
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 14, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.synapse.widget-view+json": {
       "widget_id": "12253fc4-5663-4340-9209-99de39e32625",
       "widget_type": "Synapse.DataFrame"
      },
      "text/plain": [
       "SynapseWidget(Synapse.DataFrame, 12253fc4-5663-4340-9209-99de39e32625)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_users = df_users.dropna(subset=(USER_ID_COL))\n",
    "display(df_users, summary=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1abc46c0-37a4-457b-8077-70a59119d9d6",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "Similarly, add the column named `_user_id` for later use which must only contain integers to suit your recommendation model. Therefore, use a `StringIndexer` to transform the `USER_ID_COL` into indices.\n",
    "\n",
    "**Note**: In this scenario, the book dataset already consists of an integer column named `User-ID`. However, in order to make this notebook more robust, add a `_user_id` column for compatibility with different datasets."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "46e8d227-31a2-4dcb-9e93-91f926e990c6",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:33:58.43799Z",
       "execution_start_time": "2023-09-11T14:33:55.7682439Z",
       "livy_statement_state": "available",
       "parent_msg_id": "eb33a437-9be6-4d12-9ca6-fa1c7795b136",
       "queued_time": "2023-09-11T14:33:00.6005414Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [
         {
          "completionTime": "2023-09-11T14:33:57.893GMT",
          "dataRead": 6319048,
          "dataWritten": 0,
          "description": "Job group for statement 15:\ndf_users = (\n    StringIndexer(inputCol=USER_ID_COL, outputCol=\"_user_id\")\n    .setHandleInvalid(\"skip\")\n    .fit(df_users)\n    .transform(df_users)\n    .withColumn(\"_user_id\", F.col(\"_user_id\").cast(\"int\"))\n)\n\ndisplay(df_users.sort(F.col(\"_user_id\").desc()))",
          "displayName": "getRowsInJsonString at Display.scala:452",
          "jobGroup": "15",
          "jobId": 26,
          "killedTasksSummary": {},
          "name": "getRowsInJsonString at Display.scala:452",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 3,
          "numCompletedStages": 1,
          "numCompletedTasks": 3,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 3,
          "rowCount": 29,
          "stageIds": [
           35
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:33:57.341GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:33:56.367GMT",
          "dataRead": 2081641,
          "dataWritten": 0,
          "description": "Job group for statement 15:\ndf_users = (\n    StringIndexer(inputCol=USER_ID_COL, outputCol=\"_user_id\")\n    .setHandleInvalid(\"skip\")\n    .fit(df_users)\n    .transform(df_users)\n    .withColumn(\"_user_id\", F.col(\"_user_id\").cast(\"int\"))\n)\n\ndisplay(df_users.sort(F.col(\"_user_id\").desc()))",
          "displayName": "collect at StringIndexer.scala:204",
          "jobGroup": "15",
          "jobId": 25,
          "killedTasksSummary": {},
          "name": "collect at StringIndexer.scala:204",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 1,
          "numSkippedTasks": 3,
          "numTasks": 4,
          "rowCount": 3,
          "stageIds": [
           33,
           34
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:33:56.184GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:33:56.173GMT",
          "dataRead": 6319048,
          "dataWritten": 2081641,
          "description": "Job group for statement 15:\ndf_users = (\n    StringIndexer(inputCol=USER_ID_COL, outputCol=\"_user_id\")\n    .setHandleInvalid(\"skip\")\n    .fit(df_users)\n    .transform(df_users)\n    .withColumn(\"_user_id\", F.col(\"_user_id\").cast(\"int\"))\n)\n\ndisplay(df_users.sort(F.col(\"_user_id\").desc()))",
          "displayName": "collect at StringIndexer.scala:204",
          "jobGroup": "15",
          "jobId": 24,
          "killedTasksSummary": {},
          "name": "collect at StringIndexer.scala:204",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 3,
          "numCompletedStages": 1,
          "numCompletedTasks": 3,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 3,
          "rowCount": 32,
          "stageIds": [
           32
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:33:55.868GMT",
          "usageDescription": ""
         }
        ],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 3,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 15
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 15, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.synapse.widget-view+json": {
       "widget_id": "ec85c5a8-7254-4061-9cc1-0770335cee97",
       "widget_type": "Synapse.DataFrame"
      },
      "text/plain": [
       "SynapseWidget(Synapse.DataFrame, ec85c5a8-7254-4061-9cc1-0770335cee97)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_users = (\n",
    "    StringIndexer(inputCol=USER_ID_COL, outputCol=\"_user_id\")\n",
    "    .setHandleInvalid(\"skip\")\n",
    "    .fit(df_users)\n",
    "    .transform(df_users)\n",
    "    .withColumn(\"_user_id\", F.col(\"_user_id\").cast(\"int\"))\n",
    ")\n",
    "\n",
    "display(df_users.sort(F.col(\"_user_id\").desc()))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "87b848d8-6056-43bd-8e29-a2e5381224bc",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "### Explore the Rating dataset\n",
    "Take a look at the dataframe that stores the rating data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "57ec6bae-93db-4db0-892a-9be3bc1ccf7d",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:34:01.6262948Z",
       "execution_start_time": "2023-09-11T14:33:58.9125843Z",
       "livy_statement_state": "available",
       "parent_msg_id": "1addde62-4529-474f-905d-f61d0baa3ed7",
       "queued_time": "2023-09-11T14:33:00.6043958Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [
         {
          "completionTime": "2023-09-11T14:34:01.018GMT",
          "dataRead": 1420735,
          "dataWritten": 0,
          "description": "Job group for statement 16:\ndisplay(df_ratings, summary=True)",
          "displayName": "collect at AnalysisRunner.scala:328",
          "jobGroup": "16",
          "jobId": 28,
          "killedTasksSummary": {},
          "name": "collect at AnalysisRunner.scala:328",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 1,
          "numSkippedTasks": 6,
          "numTasks": 7,
          "rowCount": 6,
          "stageIds": [
           37,
           38
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:34:00.876GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:34:00.859GMT",
          "dataRead": 22961572,
          "dataWritten": 1420735,
          "description": "Job group for statement 16:\ndisplay(df_ratings, summary=True)",
          "displayName": "collect at AnalysisRunner.scala:328",
          "jobGroup": "16",
          "jobId": 27,
          "killedTasksSummary": {},
          "name": "collect at AnalysisRunner.scala:328",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 6,
          "numCompletedStages": 1,
          "numCompletedTasks": 6,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 6,
          "rowCount": 1149786,
          "stageIds": [
           36
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:33:59.056GMT",
          "usageDescription": ""
         }
        ],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 2,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 16
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 16, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.synapse.widget-view+json": {
       "widget_id": "3f171886-f708-4d35-aea0-556e02b841dc",
       "widget_type": "Synapse.DataFrame"
      },
      "text/plain": [
       "SynapseWidget(Synapse.DataFrame, 3f171886-f708-4d35-aea0-556e02b841dc)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "display(df_ratings, summary=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2a07c665-94df-42bd-a285-ffa2b706ce4d",
   "metadata": {},
   "source": [
    "Collect the distinct ratings and save them to a list named `ratings` for later use."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b22adf2a-2b2a-4625-a4ae-aae616b166d5",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:34:03.0195913Z",
       "execution_start_time": "2023-09-11T14:34:02.0669166Z",
       "livy_statement_state": "available",
       "parent_msg_id": "4091660b-01b2-4f90-a033-2bb90a28eda6",
       "queued_time": "2023-09-11T14:33:00.6050361Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [
         {
          "completionTime": "2023-09-11T14:34:02.505GMT",
          "dataRead": 3876,
          "dataWritten": 0,
          "description": "Job group for statement 17:\nratings = [i[0] for i in df_ratings.select(RATING_COL).distinct().collect()]\nprint(ratings)",
          "displayName": "collect at /tmp/ipykernel_9996/2452077145.py:1",
          "jobGroup": "17",
          "jobId": 30,
          "killedTasksSummary": {},
          "name": "collect at /tmp/ipykernel_9996/2452077145.py:1",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 1,
          "numSkippedTasks": 6,
          "numTasks": 7,
          "rowCount": 66,
          "stageIds": [
           40,
           41
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:34:02.461GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:34:02.412GMT",
          "dataRead": 18171584,
          "dataWritten": 3876,
          "description": "Job group for statement 17:\nratings = [i[0] for i in df_ratings.select(RATING_COL).distinct().collect()]\nprint(ratings)",
          "displayName": "collect at /tmp/ipykernel_9996/2452077145.py:1",
          "jobGroup": "17",
          "jobId": 29,
          "killedTasksSummary": {},
          "name": "collect at /tmp/ipykernel_9996/2452077145.py:1",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 6,
          "numCompletedStages": 1,
          "numCompletedTasks": 6,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 6,
          "rowCount": 183,
          "stageIds": [
           39
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:34:02.226GMT",
          "usageDescription": ""
         }
        ],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 2,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 17
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 17, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1, 6, 3, 5, 9, 4, 8, 7, 10, 2, 0]\n"
     ]
    }
   ],
   "source": [
    "ratings = [i[0] for i in df_ratings.select(RATING_COL).distinct().collect()]\n",
    "print(ratings)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1220f032-b83a-45af-bf85-9015efe1715e",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "##### Plot the top 10 books with the highest ratings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "872775eb-9a61-49d9-a7ee-656bbfdc1ff1",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:34:06.3290862Z",
       "execution_start_time": "2023-09-11T14:34:03.6131246Z",
       "livy_statement_state": "available",
       "parent_msg_id": "0d7f2054-5640-4bdd-882b-67b6baddfdb9",
       "queued_time": "2023-09-11T14:33:00.6056427Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 0,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 18
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 18, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Top 10 books per number of ratings')"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 800x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(8,5))\n",
    "sns.countplot(y=\"Book-Title\",palette = 'Paired',data= df_books, order=df_books['Book-Title'].value_counts().index[0:10])\n",
    "plt.title(\"Top 10 books per number of ratings\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "705aa6f6-26aa-4e18-9a15-e49025dadfec",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "\"Selected Poems\" is most favorable among users according to ratings. The books \"Adventures of Huckleberry Finn\", \"The Secret Garden\", and \"Dracula\", have the same rating."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ac762b1f-e728-4d50-8633-95e2d1a41960",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "### Merge all datasets\n",
    "\n",
    "Once you've looked into all three dataframes corresponding to Book, User and Rating datasets, merge all three dataframes into one dataframe for a more comprehensive analysis."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "aa14f28b-d564-40b5-b6f9-4646e90d01a4",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:34:15.2828199Z",
       "execution_start_time": "2023-09-11T14:34:06.8029598Z",
       "livy_statement_state": "available",
       "parent_msg_id": "25a87b08-1cee-4aea-9382-a29b35ef8d86",
       "queued_time": "2023-09-11T14:33:00.6062511Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [
         {
          "completionTime": "2023-09-11T14:34:13.024GMT",
          "dataRead": 93167081,
          "dataWritten": 67965116,
          "description": "Job group for statement 19:\ndf_all = df_ratings.join(df_users, USER_ID_COL, \"inner\").join(\n    df_items, ITEM_ID_COL, \"inner\"\n)\ndf_all_columns = [\n    c for c in df_all.columns if c not in [\"_user_id\", \"_item_id\", RATING_COL]\n]\n\n# Reorders the columns to ensure that _user_id, _item_id, and Book-Rating are the first three columns\ndf_all = (\n    df_all.select([\"_user_id\", \"_item_id\", RATING_COL] + df_all_columns)\n    .withColumn(\"id\", F.monotonically_increasing_id())\n    .cache()\n)\n\ndisplay(df_all)",
          "displayName": "getRowsInJsonString at Display.scala:452",
          "jobGroup": "19",
          "jobId": 32,
          "killedTasksSummary": {},
          "name": "getRowsInJsonString at Display.scala:452",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 15,
          "numCompletedStages": 3,
          "numCompletedTasks": 15,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 15,
          "rowCount": 1308934,
          "stageIds": [
           45,
           43,
           44
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:34:09.367GMT",
          "usageDescription": ""
         }
        ],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 1,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 19
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 19, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.synapse.widget-view+json": {
       "widget_id": "017c9167-f3b2-4ee1-92db-476188a3c8f0",
       "widget_type": "Synapse.DataFrame"
      },
      "text/plain": [
       "SynapseWidget(Synapse.DataFrame, 017c9167-f3b2-4ee1-92db-476188a3c8f0)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_all = df_ratings.join(df_users, USER_ID_COL, \"inner\").join(\n",
    "    df_items, ITEM_ID_COL, \"inner\"\n",
    ")\n",
    "df_all_columns = [\n",
    "    c for c in df_all.columns if c not in [\"_user_id\", \"_item_id\", RATING_COL]\n",
    "]\n",
    "\n",
    "# Reorders the columns to ensure that _user_id, _item_id, and Book-Rating are the first three columns\n",
    "df_all = (\n",
    "    df_all.select([\"_user_id\", \"_item_id\", RATING_COL] + df_all_columns)\n",
    "    .withColumn(\"id\", F.monotonically_increasing_id())\n",
    "    .cache()\n",
    ")\n",
    "\n",
    "display(df_all)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "98c9912f-1f7c-4df3-97da-188db84dd551",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "Count and print the total number of distinct users, the total number of distinct books, and the total number of interactions."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "a6aecab5-c642-4cc4-9c97-d126485c6c04",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:34:44.2676271Z",
       "execution_start_time": "2023-09-11T14:34:15.8567633Z",
       "livy_statement_state": "available",
       "parent_msg_id": "923c9396-6cc1-4cc2-8a07-e5b5517ce6b5",
       "queued_time": "2023-09-11T14:33:04.4068931Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [
         {
          "completionTime": "2023-09-11T14:34:43.491GMT",
          "dataRead": 11800,
          "dataWritten": 0,
          "description": "Job group for statement 20:\nprint(f\"Total Users: {df_users.select('_user_id').distinct().count()}\")\nprint(f\"Total Items: {df_items.select('_item_id').distinct().count()}\")\nprint(f\"Total User-Item Interactions: {df_all.count()}\")",
          "displayName": "count at NativeMethodAccessorImpl.java:0",
          "jobGroup": "20",
          "jobId": 40,
          "killedTasksSummary": {},
          "name": "count at NativeMethodAccessorImpl.java:0",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 3,
          "numSkippedTasks": 214,
          "numTasks": 215,
          "rowCount": 200,
          "stageIds": [
           63,
           64,
           61,
           62
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:34:43.456GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:34:43.435GMT",
          "dataRead": 68201779,
          "dataWritten": 11800,
          "description": "Job group for statement 20:\nprint(f\"Total Users: {df_users.select('_user_id').distinct().count()}\")\nprint(f\"Total Items: {df_items.select('_item_id').distinct().count()}\")\nprint(f\"Total User-Item Interactions: {df_all.count()}\")",
          "displayName": "count at NativeMethodAccessorImpl.java:0",
          "jobGroup": "20",
          "jobId": 39,
          "killedTasksSummary": {},
          "name": "count at NativeMethodAccessorImpl.java:0",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 200,
          "numCompletedStages": 1,
          "numCompletedTasks": 200,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 2,
          "numSkippedTasks": 14,
          "numTasks": 214,
          "rowCount": 1296407,
          "stageIds": [
           60,
           58,
           59
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:34:19.569GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:34:19.322GMT",
          "dataRead": 58,
          "dataWritten": 0,
          "description": "Job group for statement 20:\nprint(f\"Total Users: {df_users.select('_user_id').distinct().count()}\")\nprint(f\"Total Items: {df_items.select('_item_id').distinct().count()}\")\nprint(f\"Total User-Item Interactions: {df_all.count()}\")",
          "displayName": "count at NativeMethodAccessorImpl.java:0",
          "jobGroup": "20",
          "jobId": 38,
          "killedTasksSummary": {},
          "name": "count at NativeMethodAccessorImpl.java:0",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 2,
          "numSkippedTasks": 9,
          "numTasks": 10,
          "rowCount": 1,
          "stageIds": [
           56,
           57,
           55
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:34:19.308GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:34:19.295GMT",
          "dataRead": 1538926,
          "dataWritten": 58,
          "description": "Job group for statement 20:\nprint(f\"Total Users: {df_users.select('_user_id').distinct().count()}\")\nprint(f\"Total Items: {df_items.select('_item_id').distinct().count()}\")\nprint(f\"Total User-Item Interactions: {df_all.count()}\")",
          "displayName": "count at NativeMethodAccessorImpl.java:0",
          "jobGroup": "20",
          "jobId": 37,
          "killedTasksSummary": {},
          "name": "count at NativeMethodAccessorImpl.java:0",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 1,
          "numSkippedTasks": 8,
          "numTasks": 9,
          "rowCount": 271361,
          "stageIds": [
           53,
           54
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:34:18.971GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:34:18.848GMT",
          "dataRead": 74660656,
          "dataWritten": 1538926,
          "description": "Job group for statement 20:\nprint(f\"Total Users: {df_users.select('_user_id').distinct().count()}\")\nprint(f\"Total Items: {df_items.select('_item_id').distinct().count()}\")\nprint(f\"Total User-Item Interactions: {df_all.count()}\")",
          "displayName": "count at NativeMethodAccessorImpl.java:0",
          "jobGroup": "20",
          "jobId": 36,
          "killedTasksSummary": {},
          "name": "count at NativeMethodAccessorImpl.java:0",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 8,
          "numCompletedStages": 1,
          "numCompletedTasks": 8,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 8,
          "rowCount": 271391,
          "stageIds": [
           52
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:34:17.872GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:34:17.529GMT",
          "dataRead": 59,
          "dataWritten": 0,
          "description": "Job group for statement 20:\nprint(f\"Total Users: {df_users.select('_user_id').distinct().count()}\")\nprint(f\"Total Items: {df_items.select('_item_id').distinct().count()}\")\nprint(f\"Total User-Item Interactions: {df_all.count()}\")",
          "displayName": "count at NativeMethodAccessorImpl.java:0",
          "jobGroup": "20",
          "jobId": 35,
          "killedTasksSummary": {},
          "name": "count at NativeMethodAccessorImpl.java:0",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 2,
          "numSkippedTasks": 4,
          "numTasks": 5,
          "rowCount": 1,
          "stageIds": [
           51,
           49,
           50
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:34:17.507GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:34:17.488GMT",
          "dataRead": 1602685,
          "dataWritten": 59,
          "description": "Job group for statement 20:\nprint(f\"Total Users: {df_users.select('_user_id').distinct().count()}\")\nprint(f\"Total Items: {df_items.select('_item_id').distinct().count()}\")\nprint(f\"Total User-Item Interactions: {df_all.count()}\")",
          "displayName": "count at NativeMethodAccessorImpl.java:0",
          "jobGroup": "20",
          "jobId": 34,
          "killedTasksSummary": {},
          "name": "count at NativeMethodAccessorImpl.java:0",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 1,
          "numSkippedTasks": 3,
          "numTasks": 4,
          "rowCount": 278859,
          "stageIds": [
           48,
           47
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:34:17.134GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:34:16.927GMT",
          "dataRead": 6319048,
          "dataWritten": 1602685,
          "description": "Job group for statement 20:\nprint(f\"Total Users: {df_users.select('_user_id').distinct().count()}\")\nprint(f\"Total Items: {df_items.select('_item_id').distinct().count()}\")\nprint(f\"Total User-Item Interactions: {df_all.count()}\")",
          "displayName": "count at NativeMethodAccessorImpl.java:0",
          "jobGroup": "20",
          "jobId": 33,
          "killedTasksSummary": {},
          "name": "count at NativeMethodAccessorImpl.java:0",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 3,
          "numCompletedStages": 1,
          "numCompletedTasks": 3,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 3,
          "rowCount": 278887,
          "stageIds": [
           46
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:34:16.254GMT",
          "usageDescription": ""
         }
        ],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 8,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 20
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 20, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total Users: 278858\n",
      "Total Items: 271360\n",
      "Total User-Item Interactions: 1031136\n"
     ]
    }
   ],
   "source": [
    "print(f\"Total Users: {df_users.select('_user_id').distinct().count()}\")\n",
    "print(f\"Total Items: {df_items.select('_item_id').distinct().count()}\")\n",
    "print(f\"Total User-Item Interactions: {df_all.count()}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "64ed7d2b-957e-46ed-b9da-07816573d80c",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "### Create visualizations\n",
    "\n",
    "Compute the most popular books and then display the top 10. Note that the `<topn>` popular items can be used for **\"Popular\"** or **\"Most purchased\"** recommendation section."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "777d503e-b7b9-41c0-98f8-72627f102a4a",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:35:12.9470242Z",
       "execution_start_time": "2023-09-11T14:34:44.7687185Z",
       "livy_statement_state": "available",
       "parent_msg_id": "c6daa4c9-7007-4ff3-9ad8-7a5eb12a6d6b",
       "queued_time": "2023-09-11T14:33:07.6446394Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [
         {
          "completionTime": "2023-09-11T14:35:12.301GMT",
          "dataRead": 74673534,
          "dataWritten": 12878,
          "description": "Job group for statement 21:\n# Compute top popular products\ndf_top_items = (\n    df_all.groupby([\"_item_id\"])\n    .count()\n    .join(df_items, \"_item_id\", \"inner\")\n    .sort([\"count\"], ascending=[0])\n)\n\n# Find top <topn> popular items\ntopn = 10\npd_top_items = df_top_items.limit(topn).toPandas()\npd_top_items.head(10)",
          "displayName": "toPandas at /tmp/ipykernel_9996/1865470362.py:11",
          "jobGroup": "21",
          "jobId": 43,
          "killedTasksSummary": {},
          "name": "toPandas at /tmp/ipykernel_9996/1865470362.py:11",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 9,
          "numCompletedStages": 2,
          "numCompletedTasks": 9,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 9,
          "rowCount": 191,
          "stageIds": [
           72,
           73
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:35:10.774GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:35:08.246GMT",
          "dataRead": 114371936,
          "dataWritten": 5058971,
          "description": "Job group for statement 21:\n# Compute top popular products\ndf_top_items = (\n    df_all.groupby([\"_item_id\"])\n    .count()\n    .join(df_items, \"_item_id\", \"inner\")\n    .sort([\"count\"], ascending=[0])\n)\n\n# Find top <topn> popular items\ntopn = 10\npd_top_items = df_top_items.limit(topn).toPandas()\npd_top_items.head(10)",
          "displayName": "toPandas at /tmp/ipykernel_9996/1865470362.py:11",
          "jobGroup": "21",
          "jobId": 41,
          "killedTasksSummary": {},
          "name": "toPandas at /tmp/ipykernel_9996/1865470362.py:11",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 200,
          "numCompletedStages": 1,
          "numCompletedTasks": 200,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 2,
          "numSkippedTasks": 14,
          "numTasks": 214,
          "rowCount": 270351,
          "stageIds": [
           66,
           67,
           65
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:34:45.308GMT",
          "usageDescription": ""
         }
        ],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 2,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 21
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 21, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>_item_id</th>\n",
       "      <th>count</th>\n",
       "      <th>ISBN</th>\n",
       "      <th>Book-Title</th>\n",
       "      <th>Book-Author</th>\n",
       "      <th>Year-Of-Publication</th>\n",
       "      <th>Publisher</th>\n",
       "      <th>Image-URL-S</th>\n",
       "      <th>Image-URL-M</th>\n",
       "      <th>Image-URL-L</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>216737</td>\n",
       "      <td>2502</td>\n",
       "      <td>0971880107</td>\n",
       "      <td>Wild Animus</td>\n",
       "      <td>Rich Shapero</td>\n",
       "      <td>2004</td>\n",
       "      <td>Too Far</td>\n",
       "      <td>http://images.amazon.com/images/P/0971880107.0...</td>\n",
       "      <td>http://images.amazon.com/images/P/0971880107.0...</td>\n",
       "      <td>http://images.amazon.com/images/P/0971880107.0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>38729</td>\n",
       "      <td>1295</td>\n",
       "      <td>0316666343</td>\n",
       "      <td>The Lovely Bones: A Novel</td>\n",
       "      <td>Alice Sebold</td>\n",
       "      <td>2002</td>\n",
       "      <td>Little, Brown</td>\n",
       "      <td>http://images.amazon.com/images/P/0316666343.0...</td>\n",
       "      <td>http://images.amazon.com/images/P/0316666343.0...</td>\n",
       "      <td>http://images.amazon.com/images/P/0316666343.0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>71056</td>\n",
       "      <td>883</td>\n",
       "      <td>0385504209</td>\n",
       "      <td>The Da Vinci Code</td>\n",
       "      <td>Dan Brown</td>\n",
       "      <td>2003</td>\n",
       "      <td>Doubleday</td>\n",
       "      <td>http://images.amazon.com/images/P/0385504209.0...</td>\n",
       "      <td>http://images.amazon.com/images/P/0385504209.0...</td>\n",
       "      <td>http://images.amazon.com/images/P/0385504209.0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>7378</td>\n",
       "      <td>732</td>\n",
       "      <td>0060928336</td>\n",
       "      <td>Divine Secrets of the Ya-Ya Sisterhood: A Novel</td>\n",
       "      <td>Rebecca Wells</td>\n",
       "      <td>1997</td>\n",
       "      <td>Perennial</td>\n",
       "      <td>http://images.amazon.com/images/P/0060928336.0...</td>\n",
       "      <td>http://images.amazon.com/images/P/0060928336.0...</td>\n",
       "      <td>http://images.amazon.com/images/P/0060928336.0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>32497</td>\n",
       "      <td>723</td>\n",
       "      <td>0312195516</td>\n",
       "      <td>The Red Tent (Bestselling Backlist)</td>\n",
       "      <td>Anita Diamant</td>\n",
       "      <td>1998</td>\n",
       "      <td>Picador USA</td>\n",
       "      <td>http://images.amazon.com/images/P/0312195516.0...</td>\n",
       "      <td>http://images.amazon.com/images/P/0312195516.0...</td>\n",
       "      <td>http://images.amazon.com/images/P/0312195516.0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>87704</td>\n",
       "      <td>647</td>\n",
       "      <td>044023722X</td>\n",
       "      <td>A Painted House</td>\n",
       "      <td>John Grisham</td>\n",
       "      <td>2001</td>\n",
       "      <td>Dell Publishing Company</td>\n",
       "      <td>http://images.amazon.com/images/P/044023722X.0...</td>\n",
       "      <td>http://images.amazon.com/images/P/044023722X.0...</td>\n",
       "      <td>http://images.amazon.com/images/P/044023722X.0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>21423</td>\n",
       "      <td>615</td>\n",
       "      <td>0142001740</td>\n",
       "      <td>The Secret Life of Bees</td>\n",
       "      <td>Sue Monk Kidd</td>\n",
       "      <td>2003</td>\n",
       "      <td>Penguin Books</td>\n",
       "      <td>http://images.amazon.com/images/P/0142001740.0...</td>\n",
       "      <td>http://images.amazon.com/images/P/0142001740.0...</td>\n",
       "      <td>http://images.amazon.com/images/P/0142001740.0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>145506</td>\n",
       "      <td>614</td>\n",
       "      <td>067976402X</td>\n",
       "      <td>Snow Falling on Cedars</td>\n",
       "      <td>David Guterson</td>\n",
       "      <td>1995</td>\n",
       "      <td>Vintage Books USA</td>\n",
       "      <td>http://images.amazon.com/images/P/067976402X.0...</td>\n",
       "      <td>http://images.amazon.com/images/P/067976402X.0...</td>\n",
       "      <td>http://images.amazon.com/images/P/067976402X.0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>133573</td>\n",
       "      <td>586</td>\n",
       "      <td>0671027360</td>\n",
       "      <td>Angels &amp;amp; Demons</td>\n",
       "      <td>Dan Brown</td>\n",
       "      <td>2001</td>\n",
       "      <td>Pocket Star</td>\n",
       "      <td>http://images.amazon.com/images/P/0671027360.0...</td>\n",
       "      <td>http://images.amazon.com/images/P/0671027360.0...</td>\n",
       "      <td>http://images.amazon.com/images/P/0671027360.0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>94175</td>\n",
       "      <td>585</td>\n",
       "      <td>0446672211</td>\n",
       "      <td>Where the Heart Is (Oprah's Book Club (Paperba...</td>\n",
       "      <td>Billie Letts</td>\n",
       "      <td>1998</td>\n",
       "      <td>Warner Books</td>\n",
       "      <td>http://images.amazon.com/images/P/0446672211.0...</td>\n",
       "      <td>http://images.amazon.com/images/P/0446672211.0...</td>\n",
       "      <td>http://images.amazon.com/images/P/0446672211.0...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   _item_id  count        ISBN  \\\n",
       "0    216737   2502  0971880107   \n",
       "1     38729   1295  0316666343   \n",
       "2     71056    883  0385504209   \n",
       "3      7378    732  0060928336   \n",
       "4     32497    723  0312195516   \n",
       "5     87704    647  044023722X   \n",
       "6     21423    615  0142001740   \n",
       "7    145506    614  067976402X   \n",
       "8    133573    586  0671027360   \n",
       "9     94175    585  0446672211   \n",
       "\n",
       "                                          Book-Title     Book-Author  \\\n",
       "0                                        Wild Animus    Rich Shapero   \n",
       "1                          The Lovely Bones: A Novel    Alice Sebold   \n",
       "2                                  The Da Vinci Code       Dan Brown   \n",
       "3    Divine Secrets of the Ya-Ya Sisterhood: A Novel   Rebecca Wells   \n",
       "4                The Red Tent (Bestselling Backlist)   Anita Diamant   \n",
       "5                                    A Painted House    John Grisham   \n",
       "6                            The Secret Life of Bees   Sue Monk Kidd   \n",
       "7                             Snow Falling on Cedars  David Guterson   \n",
       "8                                Angels &amp; Demons       Dan Brown   \n",
       "9  Where the Heart Is (Oprah's Book Club (Paperba...    Billie Letts   \n",
       "\n",
       "  Year-Of-Publication                Publisher  \\\n",
       "0                2004                  Too Far   \n",
       "1                2002            Little, Brown   \n",
       "2                2003                Doubleday   \n",
       "3                1997                Perennial   \n",
       "4                1998              Picador USA   \n",
       "5                2001  Dell Publishing Company   \n",
       "6                2003            Penguin Books   \n",
       "7                1995        Vintage Books USA   \n",
       "8                2001              Pocket Star   \n",
       "9                1998             Warner Books   \n",
       "\n",
       "                                         Image-URL-S  \\\n",
       "0  http://images.amazon.com/images/P/0971880107.0...   \n",
       "1  http://images.amazon.com/images/P/0316666343.0...   \n",
       "2  http://images.amazon.com/images/P/0385504209.0...   \n",
       "3  http://images.amazon.com/images/P/0060928336.0...   \n",
       "4  http://images.amazon.com/images/P/0312195516.0...   \n",
       "5  http://images.amazon.com/images/P/044023722X.0...   \n",
       "6  http://images.amazon.com/images/P/0142001740.0...   \n",
       "7  http://images.amazon.com/images/P/067976402X.0...   \n",
       "8  http://images.amazon.com/images/P/0671027360.0...   \n",
       "9  http://images.amazon.com/images/P/0446672211.0...   \n",
       "\n",
       "                                         Image-URL-M  \\\n",
       "0  http://images.amazon.com/images/P/0971880107.0...   \n",
       "1  http://images.amazon.com/images/P/0316666343.0...   \n",
       "2  http://images.amazon.com/images/P/0385504209.0...   \n",
       "3  http://images.amazon.com/images/P/0060928336.0...   \n",
       "4  http://images.amazon.com/images/P/0312195516.0...   \n",
       "5  http://images.amazon.com/images/P/044023722X.0...   \n",
       "6  http://images.amazon.com/images/P/0142001740.0...   \n",
       "7  http://images.amazon.com/images/P/067976402X.0...   \n",
       "8  http://images.amazon.com/images/P/0671027360.0...   \n",
       "9  http://images.amazon.com/images/P/0446672211.0...   \n",
       "\n",
       "                                         Image-URL-L  \n",
       "0  http://images.amazon.com/images/P/0971880107.0...  \n",
       "1  http://images.amazon.com/images/P/0316666343.0...  \n",
       "2  http://images.amazon.com/images/P/0385504209.0...  \n",
       "3  http://images.amazon.com/images/P/0060928336.0...  \n",
       "4  http://images.amazon.com/images/P/0312195516.0...  \n",
       "5  http://images.amazon.com/images/P/044023722X.0...  \n",
       "6  http://images.amazon.com/images/P/0142001740.0...  \n",
       "7  http://images.amazon.com/images/P/067976402X.0...  \n",
       "8  http://images.amazon.com/images/P/0671027360.0...  \n",
       "9  http://images.amazon.com/images/P/0446672211.0...  "
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Compute top popular products\n",
    "df_top_items = (\n",
    "    df_all.groupby([\"_item_id\"])\n",
    "    .count()\n",
    "    .join(df_items, \"_item_id\", \"inner\")\n",
    "    .sort([\"count\"], ascending=[0])\n",
    ")\n",
    "\n",
    "# Find top <topn> popular items\n",
    "topn = 10\n",
    "pd_top_items = df_top_items.limit(topn).toPandas()\n",
    "pd_top_items.head(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b6d7b7f5-2a26-424a-8c72-94b6ccefc41e",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:35:13.7666959Z",
       "execution_start_time": "2023-09-11T14:35:13.4311871Z",
       "livy_statement_state": "available",
       "parent_msg_id": "63980c41-c2a6-4554-947a-9241f855e488",
       "queued_time": "2023-09-11T14:33:08.5311546Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 0,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 22
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 22, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot top <topn> items\n",
    "f, ax = plt.subplots(figsize=(10, 5))\n",
    "plt.xticks(rotation=\"vertical\")\n",
    "sns.barplot(y=ITEM_INFO_COL, x=\"count\", data=pd_top_items)\n",
    "ax.tick_params(axis='x', rotation=45)\n",
    "plt.xlabel(\"Number of Ratings for the Item\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c9c82a22-f57b-4b98-8c64-1fd5e1ff1261",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "This concludes the brief analysis of the three datasets where you added unique IDs to the user and item datasets and plotted the top items."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3587d7e8-b907-4e98-b52f-15e94a7788d1",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "### Prepare data for machine learning training"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d178ac19-a670-4ec6-9825-9421c6f9d62f",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "#### Prepare training and test datasets\n",
    "\n",
    "Prior to training, you need to perform some additional data preparation steps for the ALS recommender. First cast the rating column into the correct type and then sample the training data with user ratings. Once done with data preparation, split the data into training and test datasets."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "cf38a006-521e-46cd-ac4b-ea4cf7a6fe69",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:35:14.6105286Z",
       "execution_start_time": "2023-09-11T14:35:14.2783787Z",
       "livy_statement_state": "available",
       "parent_msg_id": "0d7846cc-d933-44b0-82b7-73c8f56a778e",
       "queued_time": "2023-09-11T14:33:09.164969Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 0,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 23
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 23, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "if IS_SAMPLE:\n",
    "    # Must sort by '_user_id' before performing limit to ensure ALS work normally\n",
    "    # Note that if train and test datasets have no common _user_id, ALS will fail\n",
    "    df_all = df_all.sort(\"_user_id\").limit(SAMPLE_ROWS)\n",
    "\n",
    "# Cast column into the correct type\n",
    "df_all = df_all.withColumn(RATING_COL, F.col(RATING_COL).cast(\"float\"))\n",
    "\n",
    "# Using a fraction between 0 to 1 returns the approximate size of the dataset, i.e., 0.8 means 80% of the dataset\n",
    "# Rating = 0 means the user didn't rate the item, so it can't be used for training\n",
    "# Use the 80% if the dataset with rating > 0 as the training dataset\n",
    "fractions_train = {0: 0}\n",
    "fractions_test = {0: 0}\n",
    "for i in ratings:\n",
    "    if i == 0:\n",
    "        continue\n",
    "    fractions_train[i] = 0.8\n",
    "    fractions_test[i] = 1\n",
    "# training dataset\n",
    "train = df_all.sampleBy(RATING_COL, fractions=fractions_train)\n",
    "\n",
    "# Join with leftanti will select all rows from df_all with rating > 0 and not in train dataset, i.e., the remaining 20% of the dataset\n",
    "# test dataset\n",
    "test = df_all.join(train, on=\"id\", how=\"leftanti\").sampleBy(\n",
    "    RATING_COL, fractions=fractions_test\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0f4c12b1-050d-4404-8643-25dbe8472cc2",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "Also compute the sparsity of the dataset to gain a better understanding of the data and the problem at hand. Sparsity refers to the  situation in which feedback data is sparse and insufficient to identify similarities in users' interests."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8f90f521-bffe-4820-b9ae-45c26389e9ae",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:36:21.6524501Z",
       "execution_start_time": "2023-09-11T14:35:15.1180342Z",
       "livy_statement_state": "available",
       "parent_msg_id": "d2b54084-fcae-4318-8e5d-d66eaf0eb672",
       "queued_time": "2023-09-11T14:33:09.9427299Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [
         {
          "completionTime": "2023-09-11T14:36:19.430GMT",
          "dataRead": 122624523,
          "dataWritten": 8252587,
          "description": "Job group for statement 24:\n# Compute the sparsity of the dataset\ndef get_mat_sparsity(ratings):\n    # Count the total number of ratings in the dataset - used as numerator\n    count_nonzero = ratings.select(RATING_COL).count()\n    print(f\"Number of rows: {count_nonzero}\")\n\n    # Count the total number of distinct user_id and distinct product_id - used as denominator\n    total_elements = (\n        ratings.select(\"_user_id\").distinct().count()\n        * ratings.select(\"_item_id\").distinct().count()\n    )\n\n    # Calculate the sparsity by dividing the numerator by the denominator\n    sparsity = (1.0 - (count_nonzero * 1.0) / total_elements) * 100\n    print(\"The ratings dataframe is \", \"%.4f\" % sparsity + \"% sparse.\")\n\nget_mat_sparsity(df_all)",
          "displayName": "count at NativeMethodAccessorImpl.java:0",
          "jobGroup": "24",
          "jobId": 46,
          "killedTasksSummary": {},
          "name": "count at NativeMethodAccessorImpl.java:0",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 201,
          "numCompletedStages": 2,
          "numCompletedTasks": 201,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 2,
          "numSkippedTasks": 14,
          "numTasks": 215,
          "rowCount": 1953090,
          "stageIds": [
           84,
           85,
           82,
           83
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:35:59.237GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:35:57.866GMT",
          "dataRead": 117740557,
          "dataWritten": 3368621,
          "description": "Job group for statement 24:\n# Compute the sparsity of the dataset\ndef get_mat_sparsity(ratings):\n    # Count the total number of ratings in the dataset - used as numerator\n    count_nonzero = ratings.select(RATING_COL).count()\n    print(f\"Number of rows: {count_nonzero}\")\n\n    # Count the total number of distinct user_id and distinct product_id - used as denominator\n    total_elements = (\n        ratings.select(\"_user_id\").distinct().count()\n        * ratings.select(\"_item_id\").distinct().count()\n    )\n\n    # Calculate the sparsity by dividing the numerator by the denominator\n    sparsity = (1.0 - (count_nonzero * 1.0) / total_elements) * 100\n    print(\"The ratings dataframe is \", \"%.4f\" % sparsity + \"% sparse.\")\n\nget_mat_sparsity(df_all)",
          "displayName": "count at NativeMethodAccessorImpl.java:0",
          "jobGroup": "24",
          "jobId": 45,
          "killedTasksSummary": {},
          "name": "count at NativeMethodAccessorImpl.java:0",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 201,
          "numCompletedStages": 2,
          "numCompletedTasks": 201,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 2,
          "numSkippedTasks": 14,
          "numTasks": 215,
          "rowCount": 1953090,
          "stageIds": [
           81,
           78,
           79,
           80
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:35:37.259GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:35:35.772GMT",
          "dataRead": 117740557,
          "dataWritten": 3368621,
          "description": "Job group for statement 24:\n# Compute the sparsity of the dataset\ndef get_mat_sparsity(ratings):\n    # Count the total number of ratings in the dataset - used as numerator\n    count_nonzero = ratings.select(RATING_COL).count()\n    print(f\"Number of rows: {count_nonzero}\")\n\n    # Count the total number of distinct user_id and distinct product_id - used as denominator\n    total_elements = (\n        ratings.select(\"_user_id\").distinct().count()\n        * ratings.select(\"_item_id\").distinct().count()\n    )\n\n    # Calculate the sparsity by dividing the numerator by the denominator\n    sparsity = (1.0 - (count_nonzero * 1.0) / total_elements) * 100\n    print(\"The ratings dataframe is \", \"%.4f\" % sparsity + \"% sparse.\")\n\nget_mat_sparsity(df_all)",
          "displayName": "count at NativeMethodAccessorImpl.java:0",
          "jobGroup": "24",
          "jobId": 44,
          "killedTasksSummary": {},
          "name": "count at NativeMethodAccessorImpl.java:0",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 201,
          "numCompletedStages": 2,
          "numCompletedTasks": 201,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 2,
          "numSkippedTasks": 14,
          "numTasks": 215,
          "rowCount": 1953090,
          "stageIds": [
           74,
           75,
           76,
           77
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:35:16.115GMT",
          "usageDescription": ""
         }
        ],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 3,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 24
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 24, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of rows: 5000\n",
      "The ratings dataframe is  99.7462% sparse.\n"
     ]
    }
   ],
   "source": [
    "# Compute the sparsity of the dataset\n",
    "def get_mat_sparsity(ratings):\n",
    "    # Count the total number of ratings in the dataset - used as numerator\n",
    "    count_nonzero = ratings.select(RATING_COL).count()\n",
    "    print(f\"Number of rows: {count_nonzero}\")\n",
    "\n",
    "    # Count the total number of distinct user_id and distinct product_id - used as denominator\n",
    "    total_elements = (\n",
    "        ratings.select(\"_user_id\").distinct().count()\n",
    "        * ratings.select(\"_item_id\").distinct().count()\n",
    "    )\n",
    "\n",
    "    # Calculate the sparsity by dividing the numerator by the denominator\n",
    "    sparsity = (1.0 - (count_nonzero * 1.0) / total_elements) * 100\n",
    "    print(\"The ratings dataframe is \", \"%.4f\" % sparsity + \"% sparse.\")\n",
    "\n",
    "get_mat_sparsity(df_all)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "72adf414-50fe-42ad-9de1-e14a4f1ab5b6",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:37:06.2491974Z",
       "execution_start_time": "2023-09-11T14:36:22.1590624Z",
       "livy_statement_state": "available",
       "parent_msg_id": "b9638228-b5a2-409f-9ab3-12e522c08406",
       "queued_time": "2023-09-11T14:33:10.3749723Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [
         {
          "completionTime": "2023-09-11T14:37:04.236GMT",
          "dataRead": 122624523,
          "dataWritten": 8252587,
          "description": "Job group for statement 25:\n# Check the id range\n# Note that ALS only supports values in the integer range\nprint(f\"max user_id: {df_all.agg({'_user_id': 'max'}).collect()[0][0]}\")\nprint(f\"max item_id: {df_all.agg({'_item_id': 'max'}).collect()[0][0]}\")",
          "displayName": "collect at /tmp/ipykernel_9996/1113080625.py:4",
          "jobGroup": "25",
          "jobId": 48,
          "killedTasksSummary": {},
          "name": "collect at /tmp/ipykernel_9996/1113080625.py:4",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 201,
          "numCompletedStages": 2,
          "numCompletedTasks": 201,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 2,
          "numSkippedTasks": 14,
          "numTasks": 215,
          "rowCount": 1953090,
          "stageIds": [
           93,
           90,
           91,
           92
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:36:44.816GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:36:43.561GMT",
          "dataRead": 117740557,
          "dataWritten": 3368621,
          "description": "Job group for statement 25:\n# Check the id range\n# Note that ALS only supports values in the integer range\nprint(f\"max user_id: {df_all.agg({'_user_id': 'max'}).collect()[0][0]}\")\nprint(f\"max item_id: {df_all.agg({'_item_id': 'max'}).collect()[0][0]}\")",
          "displayName": "collect at /tmp/ipykernel_9996/1113080625.py:3",
          "jobGroup": "25",
          "jobId": 47,
          "killedTasksSummary": {},
          "name": "collect at /tmp/ipykernel_9996/1113080625.py:3",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 201,
          "numCompletedStages": 2,
          "numCompletedTasks": 201,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 2,
          "numSkippedTasks": 14,
          "numTasks": 215,
          "rowCount": 1953090,
          "stageIds": [
           88,
           89,
           86,
           87
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:36:23.284GMT",
          "usageDescription": ""
         }
        ],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 2,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 25
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 25, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "max user_id: 1347\n",
      "max item_id: 271357\n"
     ]
    }
   ],
   "source": [
    "# Check the id range\n",
    "# Note that ALS only supports values in the integer range\n",
    "print(f\"max user_id: {df_all.agg({'_user_id': 'max'}).collect()[0][0]}\")\n",
    "print(f\"max item_id: {df_all.agg({'_item_id': 'max'}).collect()[0][0]}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "dc50ef68",
   "metadata": {},
   "source": [
    "## Step 3: Develop and train the model"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2762eb40-c336-4212-8de0-e09d16b23aa4",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "With your data in place, you can now define the model. Train an Alternating Least Squares (ALS) recommender to give users personalized recommendations."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "341ddf1e-3e14-4fa5-8f2a-fea5417bf37a",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "### The machine learning model\n",
    "\n",
    "With the train and test datasets defined, you can now get started on the recommendation model. Spark ML provides a convenient API for building an ALS model. However, the model default setting does not perform well at handling data sparsity and the cold start problem (the difficulty to make recommendations when the users or the items are new).\n",
    "Combine cross validation and auto hyperparameter tuning to improve the performance of the model."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7e6fce1a-0f33-4611-9e24-d318cbd3d992",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "You need to import the required libraries for model training and evaluation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "230ccb09-ead9-46e3-a77f-fdbf848e26ce",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:37:07.017819Z",
       "execution_start_time": "2023-09-11T14:37:06.6892845Z",
       "livy_statement_state": "available",
       "parent_msg_id": "67d5c8b0-0cdf-4798-bf7f-1f65387af245",
       "queued_time": "2023-09-11T14:33:10.9244514Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 0,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 26
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 26, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Import Spark required libraries\n",
    "from pyspark.ml.evaluation import RegressionEvaluator\n",
    "from pyspark.ml.recommendation import ALS\n",
    "from pyspark.ml.tuning import ParamGridBuilder, CrossValidator, TrainValidationSplit"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "627ed994-723a-4f4f-b918-a630f7eaa58b",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:37:07.8290722Z",
       "execution_start_time": "2023-09-11T14:37:07.5055289Z",
       "livy_statement_state": "available",
       "parent_msg_id": "f478f882-8dcf-47d0-b688-9f75f396f4a1",
       "queued_time": "2023-09-11T14:33:11.4577141Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 0,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 27
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 27, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Specify the training parameters\n",
    "num_epochs = 1  # Number of epochs, use 1 to reduce the training time\n",
    "rank_size_list = [64]  # Values of rank in ALS for tuning\n",
    "reg_param_list = [0.01, 0.1]  # Values of regParam in ALS for tuning\n",
    "model_tuning_method = \"TrainValidationSplit\"  # TrainValidationSplit or CrossValidator"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c24a52b2-de98-4200-aed8-dad8913b5123",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:37:08.5970437Z",
       "execution_start_time": "2023-09-11T14:37:08.2751747Z",
       "livy_statement_state": "available",
       "parent_msg_id": "288740c9-1fc0-4ade-a4e4-5271bb85a473",
       "queued_time": "2023-09-11T14:33:12.1782096Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 0,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 28
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 28, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Build the recommendation model using ALS on the training data\n",
    "# Set the cold start strategy to 'drop' to ensure that you don't get NaN evaluation metrics\n",
    "als = ALS(\n",
    "    maxIter=num_epochs,\n",
    "    userCol=\"_user_id\",\n",
    "    itemCol=\"_item_id\",\n",
    "    ratingCol=RATING_COL,\n",
    "    coldStartStrategy=\"drop\",\n",
    "    implicitPrefs=False,\n",
    "    nonnegative=True,\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "28c7afa4-4038-4577-9852-84ba035de2fc",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "### Tune model hyperparameters\n",
    "\n",
    "In the following code, you construct a grid of parameters to search over the hyperparameters as well as a regression evaluator with root mean square error as the evaluation metric."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "77e7da60-2690-46ed-bdac-e46293c4d661",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:37:09.4210649Z",
       "execution_start_time": "2023-09-11T14:37:09.0927325Z",
       "livy_statement_state": "available",
       "parent_msg_id": "79077b3a-e9e7-409c-8863-e4723fac2731",
       "queued_time": "2023-09-11T14:33:12.7850075Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 0,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 29
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 29, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of models to be tested:  2\n"
     ]
    }
   ],
   "source": [
    "#  Construct a grid search to select the best values for the training parameters\n",
    "param_grid = (\n",
    "    ParamGridBuilder()\n",
    "    .addGrid(als.rank, rank_size_list)\n",
    "    .addGrid(als.regParam, reg_param_list)\n",
    "    .build()\n",
    ")\n",
    "\n",
    "print(\"Number of models to be tested: \", len(param_grid))\n",
    "\n",
    "# Define evaluator and set the loss fucntion to root mean squared error (RMSE) \n",
    "evaluator = RegressionEvaluator(\n",
    "    metricName=\"rmse\", labelCol=RATING_COL, predictionCol=\"prediction\"\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e873316a-af0b-4899-8719-896c87e9cf18",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "Initiate different model tuning methods based on the pre-configured parameters. Further information of model tuning can be found at [Spark ML tuning](https://spark.apache.org/docs/latest/ml-tuning.html)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d06c1ff2-244d-488b-93fd-72ddc661d203",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:37:10.2621997Z",
       "execution_start_time": "2023-09-11T14:37:09.9178493Z",
       "livy_statement_state": "available",
       "parent_msg_id": "448ff2b5-9da8-4843-9f2b-42bd108e8727",
       "queued_time": "2023-09-11T14:33:13.5780643Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 0,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 30
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 30, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Build cross validation using CrossValidator and TrainValidationSplit\n",
    "if model_tuning_method == \"CrossValidator\":\n",
    "    tuner = CrossValidator(\n",
    "        estimator=als,\n",
    "        estimatorParamMaps=param_grid,\n",
    "        evaluator=evaluator,\n",
    "        numFolds=5,\n",
    "        collectSubModels=True,\n",
    "    )\n",
    "elif model_tuning_method == \"TrainValidationSplit\":\n",
    "    tuner = TrainValidationSplit(\n",
    "        estimator=als,\n",
    "        estimatorParamMaps=param_grid,\n",
    "        evaluator=evaluator,\n",
    "        # 80% of the training data will be used for training, 20% for validation\n",
    "        trainRatio=0.8,\n",
    "        collectSubModels=True,\n",
    "    )\n",
    "else:\n",
    "    raise ValueError(f\"Unknown model_tuning_method: {model_tuning_method}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f19bbe76-3dbc-411a-8bd4-9450ab4f822e",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "### Evaluate the model\n",
    "\n",
    "You now have different models to compare by evaluating them on the test data. If a model has been well trained, it should have high metrics on the datasets.\n",
    "\n",
    "If the model is overfitted, you may need to increase the size of the training data or reduce some of the redundant features. You may also need to change the model's architecture or fine-tune its hyperparameters.\n",
    "\n",
    "Note that if the R-Squared metric value is negative, it indicates that the trained model performs worse than a horizontal straight line, suggesting that the data is not explained by the trained model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "298f3031-73db-4e24-875d-5bae32871819",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:37:11.0502326Z",
       "execution_start_time": "2023-09-11T14:37:10.7110836Z",
       "livy_statement_state": "available",
       "parent_msg_id": "78e296f0-8d9c-4cf8-969c-03b9f78ad272",
       "queued_time": "2023-09-11T14:33:14.1980785Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 0,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 31
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 31, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def evaluate(model, data, verbose=0):\n",
    "    \"\"\"\n",
    "    Evaluate the model by computing rmse, mae, r2 and variance over the data.\n",
    "    \"\"\"\n",
    "\n",
    "    predictions = model.transform(data).withColumn(\n",
    "        \"prediction\", F.col(\"prediction\").cast(\"double\")\n",
    "    )\n",
    "\n",
    "    if verbose > 1:\n",
    "        # Show 10 predictions\n",
    "        predictions.select(\"_user_id\", \"_item_id\", RATING_COL, \"prediction\").limit(\n",
    "            10\n",
    "        ).show()\n",
    "\n",
    "    # Initialize the regression evaluator\n",
    "    evaluator = RegressionEvaluator(predictionCol=\"prediction\", labelCol=RATING_COL)\n",
    "\n",
    "    _evaluator = lambda metric: evaluator.setMetricName(metric).evaluate(predictions)\n",
    "    rmse = _evaluator(\"rmse\")\n",
    "    mae = _evaluator(\"mae\")\n",
    "    r2 = _evaluator(\"r2\")\n",
    "    var = _evaluator(\"var\")\n",
    "\n",
    "    if verbose > 0:\n",
    "        print(f\"RMSE score = {rmse}\")\n",
    "        print(f\"MAE score = {mae}\")\n",
    "        print(f\"R2 score = {r2}\")\n",
    "        print(f\"Explained variance = {var}\")\n",
    "\n",
    "    return predictions, (rmse, mae, r2, var)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9f0ed2ed-6bf2-4a92-b395-e938422b526f",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "### Track the experiment with MLflow\n",
    "\n",
    "Start the training and evaluation, then use MLflow to track all experiments and log parameters, metrics, and the models."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "7e2d917f-b9d2-4336-8c36-f1bfae30bade",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:41:29.9567152Z",
       "execution_start_time": "2023-09-11T14:37:11.5053384Z",
       "livy_statement_state": "available",
       "parent_msg_id": "bce8e474-32ce-43bd-aef0-ee7337db6034",
       "queued_time": "2023-09-11T14:33:15.1700944Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [
         {
          "completionTime": "2023-09-11T14:38:12.760GMT",
          "dataRead": 7690,
          "dataWritten": 0,
          "description": "Job group for statement 32:\nfrom mlflow.models.signature import infer_signature\n\nwith mlflow.start_run(run_name=\"als\"):\n    # Train models\n    models = tuner.fit(train)\n    best_metrics = {\"RMSE\": 10e6, \"MAE\": 10e6, \"R2\": 0, \"Explained variance\": 0}\n    best_index = 0\n    # Evaluate models\n    # Log model, metrics and parameters\n    for idx, model in enumerate(models.subModels):\n        with mlflow.start_run(nested=True, run_name=f\"als_{idx}\") as run:\n            print(\"\nEvaluating on testing data:\")\n            print(f\"subModel No. {idx + 1}\")\n            predictions, (rmse, mae, r2, var) = evaluate(model, test, verbose=1)\n\n            signature = infer_signature(\n                train.select([\"_user_id\", \"_item_id\"]),\n                predictions.select([\"_user_id\", \"_item_id\", \"prediction\"]),\n            )\n            print(\"log model:\")\n            mlflow.spark.log_model(\n                model,\n                f\"{EXPERIMENT_NAME}-alsmodel\",\n                signature=signature,\n                registered_model_name=f\"{EXPERIMENT_NAME...",
          "displayName": "treeAggregate at Statistics.scala:58",
          "jobGroup": "32",
          "jobId": 70,
          "killedTasksSummary": {},
          "name": "treeAggregate at Statistics.scala:58",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 4,
          "numSkippedTasks": 215,
          "numTasks": 216,
          "rowCount": 307,
          "stageIds": [
           201,
           202,
           203,
           199,
           200
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:38:11.987GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:38:11.752GMT",
          "dataRead": 222442,
          "dataWritten": 295874,
          "description": "Job group for statement 32:\nfrom mlflow.models.signature import infer_signature\n\nwith mlflow.start_run(run_name=\"als\"):\n    # Train models\n    models = tuner.fit(train)\n    best_metrics = {\"RMSE\": 10e6, \"MAE\": 10e6, \"R2\": 0, \"Explained variance\": 0}\n    best_index = 0\n    # Evaluate models\n    # Log model, metrics and parameters\n    for idx, model in enumerate(models.subModels):\n        with mlflow.start_run(nested=True, run_name=f\"als_{idx}\") as run:\n            print(\"\nEvaluating on testing data:\")\n            print(f\"subModel No. {idx + 1}\")\n            predictions, (rmse, mae, r2, var) = evaluate(model, test, verbose=1)\n\n            signature = infer_signature(\n                train.select([\"_user_id\", \"_item_id\"]),\n                predictions.select([\"_user_id\", \"_item_id\", \"prediction\"]),\n            )\n            print(\"log model:\")\n            mlflow.spark.log_model(\n                model,\n                f\"{EXPERIMENT_NAME}-alsmodel\",\n                signature=signature,\n                registered_model_name=f\"{EXPERIMENT_NAME...",
          "displayName": "rdd at RegressionEvaluator.scala:125",
          "jobGroup": "32",
          "jobId": 67,
          "killedTasksSummary": {},
          "name": "rdd at RegressionEvaluator.scala:125",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 10,
          "numCompletedStages": 1,
          "numCompletedTasks": 10,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 7,
          "numSkippedTasks": 227,
          "numTasks": 237,
          "rowCount": 1748,
          "stageIds": [
           186,
           183,
           190,
           187,
           188,
           189,
           193,
           182
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:38:06.138GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:38:11.686GMT",
          "dataRead": 177403,
          "dataWritten": 60706,
          "description": "Job group for statement 32:\nfrom mlflow.models.signature import infer_signature\n\nwith mlflow.start_run(run_name=\"als\"):\n    # Train models\n    models = tuner.fit(train)\n    best_metrics = {\"RMSE\": 10e6, \"MAE\": 10e6, \"R2\": 0, \"Explained variance\": 0}\n    best_index = 0\n    # Evaluate models\n    # Log model, metrics and parameters\n    for idx, model in enumerate(models.subModels):\n        with mlflow.start_run(nested=True, run_name=f\"als_{idx}\") as run:\n            print(\"\nEvaluating on testing data:\")\n            print(f\"subModel No. {idx + 1}\")\n            predictions, (rmse, mae, r2, var) = evaluate(model, test, verbose=1)\n\n            signature = infer_signature(\n                train.select([\"_user_id\", \"_item_id\"]),\n                predictions.select([\"_user_id\", \"_item_id\", \"prediction\"]),\n            )\n            print(\"log model:\")\n            mlflow.spark.log_model(\n                model,\n                f\"{EXPERIMENT_NAME}-alsmodel\",\n                signature=signature,\n                registered_model_name=f\"{EXPERIMENT_NAME...",
          "displayName": "rdd at RegressionEvaluator.scala:125",
          "jobGroup": "32",
          "jobId": 66,
          "killedTasksSummary": {},
          "name": "rdd at RegressionEvaluator.scala:125",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 10,
          "numCompletedStages": 1,
          "numCompletedTasks": 10,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 6,
          "numSkippedTasks": 223,
          "numTasks": 247,
          "rowCount": 521,
          "stageIds": [
           186,
           183,
           190,
           187,
           191,
           188,
           192,
           189,
           182
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:38:05.869GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:38:06.491GMT",
          "dataRead": 38015,
          "dataWritten": 7690,
          "description": "Job group for statement 32:\nfrom mlflow.models.signature import infer_signature\n\nwith mlflow.start_run(run_name=\"als\"):\n    # Train models\n    models = tuner.fit(train)\n    best_metrics = {\"RMSE\": 10e6, \"MAE\": 10e6, \"R2\": 0, \"Explained variance\": 0}\n    best_index = 0\n    # Evaluate models\n    # Log model, metrics and parameters\n    for idx, model in enumerate(models.subModels):\n        with mlflow.start_run(nested=True, run_name=f\"als_{idx}\") as run:\n            print(\"\nEvaluating on testing data:\")\n            print(f\"subModel No. {idx + 1}\")\n            predictions, (rmse, mae, r2, var) = evaluate(model, test, verbose=1)\n\n            signature = infer_signature(\n                train.select([\"_user_id\", \"_item_id\"]),\n                predictions.select([\"_user_id\", \"_item_id\", \"prediction\"]),\n            )\n            print(\"log model:\")\n            mlflow.spark.log_model(\n                model,\n                f\"{EXPERIMENT_NAME}-alsmodel\",\n                signature=signature,\n                registered_model_name=f\"{EXPERIMENT_NAME...",
          "displayName": "rdd at RegressionEvaluator.scala:125",
          "jobGroup": "32",
          "jobId": 65,
          "killedTasksSummary": {},
          "name": "rdd at RegressionEvaluator.scala:125",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 3,
          "numSkippedTasks": 214,
          "numTasks": 215,
          "rowCount": 308,
          "stageIds": [
           183,
           184,
           185,
           182
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:38:05.599GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:38:05.443GMT",
          "dataRead": 200332,
          "dataWritten": 0,
          "description": "Job group for statement 32:\nfrom mlflow.models.signature import infer_signature\n\nwith mlflow.start_run(run_name=\"als\"):\n    # Train models\n    models = tuner.fit(train)\n    best_metrics = {\"RMSE\": 10e6, \"MAE\": 10e6, \"R2\": 0, \"Explained variance\": 0}\n    best_index = 0\n    # Evaluate models\n    # Log model, metrics and parameters\n    for idx, model in enumerate(models.subModels):\n        with mlflow.start_run(nested=True, run_name=f\"als_{idx}\") as run:\n            print(\"\nEvaluating on testing data:\")\n            print(f\"subModel No. {idx + 1}\")\n            predictions, (rmse, mae, r2, var) = evaluate(model, test, verbose=1)\n\n            signature = infer_signature(\n                train.select([\"_user_id\", \"_item_id\"]),\n                predictions.select([\"_user_id\", \"_item_id\", \"prediction\"]),\n            )\n            print(\"log model:\")\n            mlflow.spark.log_model(\n                model,\n                f\"{EXPERIMENT_NAME}-alsmodel\",\n                signature=signature,\n                registered_model_name=f\"{EXPERIMENT_NAME...",
          "displayName": "count at ALS.scala:1080",
          "jobGroup": "32",
          "jobId": 64,
          "killedTasksSummary": {},
          "name": "count at ALS.scala:1080",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 10,
          "numCompletedStages": 1,
          "numCompletedTasks": 10,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 7,
          "numSkippedTasks": 227,
          "numTasks": 237,
          "rowCount": 120,
          "stageIds": [
           179,
           180,
           176,
           177,
           174,
           181,
           178,
           175
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:38:03.784GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:38:03.766GMT",
          "dataRead": 412493,
          "dataWritten": 328197,
          "description": "Job group for statement 32:\nfrom mlflow.models.signature import infer_signature\n\nwith mlflow.start_run(run_name=\"als\"):\n    # Train models\n    models = tuner.fit(train)\n    best_metrics = {\"RMSE\": 10e6, \"MAE\": 10e6, \"R2\": 0, \"Explained variance\": 0}\n    best_index = 0\n    # Evaluate models\n    # Log model, metrics and parameters\n    for idx, model in enumerate(models.subModels):\n        with mlflow.start_run(nested=True, run_name=f\"als_{idx}\") as run:\n            print(\"\nEvaluating on testing data:\")\n            print(f\"subModel No. {idx + 1}\")\n            predictions, (rmse, mae, r2, var) = evaluate(model, test, verbose=1)\n\n            signature = infer_signature(\n                train.select([\"_user_id\", \"_item_id\"]),\n                predictions.select([\"_user_id\", \"_item_id\", \"prediction\"]),\n            )\n            print(\"log model:\")\n            mlflow.spark.log_model(\n                model,\n                f\"{EXPERIMENT_NAME}-alsmodel\",\n                signature=signature,\n                registered_model_name=f\"{EXPERIMENT_NAME...",
          "displayName": "count at ALS.scala:1075",
          "jobGroup": "32",
          "jobId": 63,
          "killedTasksSummary": {},
          "name": "count at ALS.scala:1075",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 30,
          "numCompletedStages": 3,
          "numCompletedTasks": 30,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 6,
          "numSkippedTasks": 217,
          "numTasks": 247,
          "rowCount": 460,
          "stageIds": [
           171,
           168,
           165,
           172,
           169,
           166,
           173,
           170,
           167
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:38:00.256GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:38:00.220GMT",
          "dataRead": 45002,
          "dataWritten": 9386,
          "description": "Job group for statement 32:\nfrom mlflow.models.signature import infer_signature\n\nwith mlflow.start_run(run_name=\"als\"):\n    # Train models\n    models = tuner.fit(train)\n    best_metrics = {\"RMSE\": 10e6, \"MAE\": 10e6, \"R2\": 0, \"Explained variance\": 0}\n    best_index = 0\n    # Evaluate models\n    # Log model, metrics and parameters\n    for idx, model in enumerate(models.subModels):\n        with mlflow.start_run(nested=True, run_name=f\"als_{idx}\") as run:\n            print(\"\nEvaluating on testing data:\")\n            print(f\"subModel No. {idx + 1}\")\n            predictions, (rmse, mae, r2, var) = evaluate(model, test, verbose=1)\n\n            signature = infer_signature(\n                train.select([\"_user_id\", \"_item_id\"]),\n                predictions.select([\"_user_id\", \"_item_id\", \"prediction\"]),\n            )\n            print(\"log model:\")\n            mlflow.spark.log_model(\n                model,\n                f\"{EXPERIMENT_NAME}-alsmodel\",\n                signature=signature,\n                registered_model_name=f\"{EXPERIMENT_NAME...",
          "displayName": "count at ALS.scala:980",
          "jobGroup": "32",
          "jobId": 62,
          "killedTasksSummary": {},
          "name": "count at ALS.scala:980",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 11,
          "numCompletedStages": 2,
          "numCompletedTasks": 11,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 4,
          "numSkippedTasks": 215,
          "numTasks": 226,
          "rowCount": 300,
          "stageIds": [
           161,
           162,
           159,
           163,
           160,
           164
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:37:57.752GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:37:57.504GMT",
          "dataRead": 427173,
          "dataWritten": 17077,
          "description": "Job group for statement 32:\nfrom mlflow.models.signature import infer_signature\n\nwith mlflow.start_run(run_name=\"als\"):\n    # Train models\n    models = tuner.fit(train)\n    best_metrics = {\"RMSE\": 10e6, \"MAE\": 10e6, \"R2\": 0, \"Explained variance\": 0}\n    best_index = 0\n    # Evaluate models\n    # Log model, metrics and parameters\n    for idx, model in enumerate(models.subModels):\n        with mlflow.start_run(nested=True, run_name=f\"als_{idx}\") as run:\n            print(\"\nEvaluating on testing data:\")\n            print(f\"subModel No. {idx + 1}\")\n            predictions, (rmse, mae, r2, var) = evaluate(model, test, verbose=1)\n\n            signature = infer_signature(\n                train.select([\"_user_id\", \"_item_id\"]),\n                predictions.select([\"_user_id\", \"_item_id\", \"prediction\"]),\n            )\n            print(\"log model:\")\n            mlflow.spark.log_model(\n                model,\n                f\"{EXPERIMENT_NAME}-alsmodel\",\n                signature=signature,\n                registered_model_name=f\"{EXPERIMENT_NAME...",
          "displayName": "count at ALS.scala:973",
          "jobGroup": "32",
          "jobId": 61,
          "killedTasksSummary": {},
          "name": "count at ALS.scala:973",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 12,
          "numCompletedStages": 3,
          "numCompletedTasks": 12,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 3,
          "numSkippedTasks": 214,
          "numTasks": 226,
          "rowCount": 401,
          "stageIds": [
           153,
           157,
           154,
           158,
           155,
           156
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:37:55.247GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:37:54.773GMT",
          "dataRead": 410096,
          "dataWritten": 0,
          "description": "Job group for statement 32:\nfrom mlflow.models.signature import infer_signature\n\nwith mlflow.start_run(run_name=\"als\"):\n    # Train models\n    models = tuner.fit(train)\n    best_metrics = {\"RMSE\": 10e6, \"MAE\": 10e6, \"R2\": 0, \"Explained variance\": 0}\n    best_index = 0\n    # Evaluate models\n    # Log model, metrics and parameters\n    for idx, model in enumerate(models.subModels):\n        with mlflow.start_run(nested=True, run_name=f\"als_{idx}\") as run:\n            print(\"\nEvaluating on testing data:\")\n            print(f\"subModel No. {idx + 1}\")\n            predictions, (rmse, mae, r2, var) = evaluate(model, test, verbose=1)\n\n            signature = infer_signature(\n                train.select([\"_user_id\", \"_item_id\"]),\n                predictions.select([\"_user_id\", \"_item_id\", \"prediction\"]),\n            )\n            print(\"log model:\")\n            mlflow.spark.log_model(\n                model,\n                f\"{EXPERIMENT_NAME}-alsmodel\",\n                signature=signature,\n                registered_model_name=f\"{EXPERIMENT_NAME...",
          "displayName": "isEmpty at ALS.scala:960",
          "jobGroup": "32",
          "jobId": 60,
          "killedTasksSummary": {},
          "name": "isEmpty at ALS.scala:960",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 3,
          "numSkippedTasks": 214,
          "numTasks": 215,
          "rowCount": 1,
          "stageIds": [
           150,
           151,
           152,
           149
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:37:54.271GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:37:53.649GMT",
          "dataRead": 7690,
          "dataWritten": 0,
          "description": "Job group for statement 32:\nfrom mlflow.models.signature import infer_signature\n\nwith mlflow.start_run(run_name=\"als\"):\n    # Train models\n    models = tuner.fit(train)\n    best_metrics = {\"RMSE\": 10e6, \"MAE\": 10e6, \"R2\": 0, \"Explained variance\": 0}\n    best_index = 0\n    # Evaluate models\n    # Log model, metrics and parameters\n    for idx, model in enumerate(models.subModels):\n        with mlflow.start_run(nested=True, run_name=f\"als_{idx}\") as run:\n            print(\"\nEvaluating on testing data:\")\n            print(f\"subModel No. {idx + 1}\")\n            predictions, (rmse, mae, r2, var) = evaluate(model, test, verbose=1)\n\n            signature = infer_signature(\n                train.select([\"_user_id\", \"_item_id\"]),\n                predictions.select([\"_user_id\", \"_item_id\", \"prediction\"]),\n            )\n            print(\"log model:\")\n            mlflow.spark.log_model(\n                model,\n                f\"{EXPERIMENT_NAME}-alsmodel\",\n                signature=signature,\n                registered_model_name=f\"{EXPERIMENT_NAME...",
          "displayName": "treeAggregate at Statistics.scala:58",
          "jobGroup": "32",
          "jobId": 59,
          "killedTasksSummary": {},
          "name": "treeAggregate at Statistics.scala:58",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 4,
          "numSkippedTasks": 215,
          "numTasks": 216,
          "rowCount": 307,
          "stageIds": [
           147,
           144,
           148,
           145,
           146
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:37:53.000GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:37:51.679GMT",
          "dataRead": 353666,
          "dataWritten": 295948,
          "description": "Job group for statement 32:\nfrom mlflow.models.signature import infer_signature\n\nwith mlflow.start_run(run_name=\"als\"):\n    # Train models\n    models = tuner.fit(train)\n    best_metrics = {\"RMSE\": 10e6, \"MAE\": 10e6, \"R2\": 0, \"Explained variance\": 0}\n    best_index = 0\n    # Evaluate models\n    # Log model, metrics and parameters\n    for idx, model in enumerate(models.subModels):\n        with mlflow.start_run(nested=True, run_name=f\"als_{idx}\") as run:\n            print(\"\nEvaluating on testing data:\")\n            print(f\"subModel No. {idx + 1}\")\n            predictions, (rmse, mae, r2, var) = evaluate(model, test, verbose=1)\n\n            signature = infer_signature(\n                train.select([\"_user_id\", \"_item_id\"]),\n                predictions.select([\"_user_id\", \"_item_id\", \"prediction\"]),\n            )\n            print(\"log model:\")\n            mlflow.spark.log_model(\n                model,\n                f\"{EXPERIMENT_NAME}-alsmodel\",\n                signature=signature,\n                registered_model_name=f\"{EXPERIMENT_NAME...",
          "displayName": "rdd at RegressionEvaluator.scala:125",
          "jobGroup": "32",
          "jobId": 56,
          "killedTasksSummary": {},
          "name": "rdd at RegressionEvaluator.scala:125",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 10,
          "numCompletedStages": 1,
          "numCompletedTasks": 10,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 7,
          "numSkippedTasks": 227,
          "numTasks": 237,
          "rowCount": 2542,
          "stageIds": [
           135,
           132,
           133,
           138,
           127,
           134,
           131,
           128
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:37:47.158GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:37:51.629GMT",
          "dataRead": 59673,
          "dataWritten": 59633,
          "description": "Job group for statement 32:\nfrom mlflow.models.signature import infer_signature\n\nwith mlflow.start_run(run_name=\"als\"):\n    # Train models\n    models = tuner.fit(train)\n    best_metrics = {\"RMSE\": 10e6, \"MAE\": 10e6, \"R2\": 0, \"Explained variance\": 0}\n    best_index = 0\n    # Evaluate models\n    # Log model, metrics and parameters\n    for idx, model in enumerate(models.subModels):\n        with mlflow.start_run(nested=True, run_name=f\"als_{idx}\") as run:\n            print(\"\nEvaluating on testing data:\")\n            print(f\"subModel No. {idx + 1}\")\n            predictions, (rmse, mae, r2, var) = evaluate(model, test, verbose=1)\n\n            signature = infer_signature(\n                train.select([\"_user_id\", \"_item_id\"]),\n                predictions.select([\"_user_id\", \"_item_id\", \"prediction\"]),\n            )\n            print(\"log model:\")\n            mlflow.spark.log_model(\n                model,\n                f\"{EXPERIMENT_NAME}-alsmodel\",\n                signature=signature,\n                registered_model_name=f\"{EXPERIMENT_NAME...",
          "displayName": "rdd at RegressionEvaluator.scala:125",
          "jobGroup": "32",
          "jobId": 55,
          "killedTasksSummary": {},
          "name": "rdd at RegressionEvaluator.scala:125",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 10,
          "numCompletedStages": 1,
          "numCompletedTasks": 10,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 8,
          "numSkippedTasks": 237,
          "numTasks": 247,
          "rowCount": 482,
          "stageIds": [
           135,
           132,
           136,
           133,
           137,
           127,
           134,
           131,
           128
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:37:46.665GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:37:52.944GMT",
          "dataRead": 180922708,
          "dataWritten": 124534845,
          "description": "Job group for statement 32:\nfrom mlflow.models.signature import infer_signature\n\nwith mlflow.start_run(run_name=\"als\"):\n    # Train models\n    models = tuner.fit(train)\n    best_metrics = {\"RMSE\": 10e6, \"MAE\": 10e6, \"R2\": 0, \"Explained variance\": 0}\n    best_index = 0\n    # Evaluate models\n    # Log model, metrics and parameters\n    for idx, model in enumerate(models.subModels):\n        with mlflow.start_run(nested=True, run_name=f\"als_{idx}\") as run:\n            print(\"\nEvaluating on testing data:\")\n            print(f\"subModel No. {idx + 1}\")\n            predictions, (rmse, mae, r2, var) = evaluate(model, test, verbose=1)\n\n            signature = infer_signature(\n                train.select([\"_user_id\", \"_item_id\"]),\n                predictions.select([\"_user_id\", \"_item_id\", \"prediction\"]),\n            )\n            print(\"log model:\")\n            mlflow.spark.log_model(\n                model,\n                f\"{EXPERIMENT_NAME}-alsmodel\",\n                signature=signature,\n                registered_model_name=f\"{EXPERIMENT_NAME...",
          "displayName": "rdd at RegressionEvaluator.scala:125",
          "jobGroup": "32",
          "jobId": 54,
          "killedTasksSummary": {},
          "name": "rdd at RegressionEvaluator.scala:125",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 201,
          "numCompletedStages": 2,
          "numCompletedTasks": 201,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 215,
          "rowCount": 1953397,
          "stageIds": [
           129,
           130,
           127,
           128
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:37:46.292GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:37:45.621GMT",
          "dataRead": 200332,
          "dataWritten": 0,
          "description": "Job group for statement 32:\nfrom mlflow.models.signature import infer_signature\n\nwith mlflow.start_run(run_name=\"als\"):\n    # Train models\n    models = tuner.fit(train)\n    best_metrics = {\"RMSE\": 10e6, \"MAE\": 10e6, \"R2\": 0, \"Explained variance\": 0}\n    best_index = 0\n    # Evaluate models\n    # Log model, metrics and parameters\n    for idx, model in enumerate(models.subModels):\n        with mlflow.start_run(nested=True, run_name=f\"als_{idx}\") as run:\n            print(\"\nEvaluating on testing data:\")\n            print(f\"subModel No. {idx + 1}\")\n            predictions, (rmse, mae, r2, var) = evaluate(model, test, verbose=1)\n\n            signature = infer_signature(\n                train.select([\"_user_id\", \"_item_id\"]),\n                predictions.select([\"_user_id\", \"_item_id\", \"prediction\"]),\n            )\n            print(\"log model:\")\n            mlflow.spark.log_model(\n                model,\n                f\"{EXPERIMENT_NAME}-alsmodel\",\n                signature=signature,\n                registered_model_name=f\"{EXPERIMENT_NAME...",
          "displayName": "count at ALS.scala:1080",
          "jobGroup": "32",
          "jobId": 53,
          "killedTasksSummary": {},
          "name": "count at ALS.scala:1080",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 10,
          "numCompletedStages": 1,
          "numCompletedTasks": 10,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 7,
          "numSkippedTasks": 227,
          "numTasks": 237,
          "rowCount": 120,
          "stageIds": [
           121,
           125,
           122,
           119,
           126,
           123,
           120,
           124
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:37:44.820GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:37:44.808GMT",
          "dataRead": 413008,
          "dataWritten": 328712,
          "description": "Job group for statement 32:\nfrom mlflow.models.signature import infer_signature\n\nwith mlflow.start_run(run_name=\"als\"):\n    # Train models\n    models = tuner.fit(train)\n    best_metrics = {\"RMSE\": 10e6, \"MAE\": 10e6, \"R2\": 0, \"Explained variance\": 0}\n    best_index = 0\n    # Evaluate models\n    # Log model, metrics and parameters\n    for idx, model in enumerate(models.subModels):\n        with mlflow.start_run(nested=True, run_name=f\"als_{idx}\") as run:\n            print(\"\nEvaluating on testing data:\")\n            print(f\"subModel No. {idx + 1}\")\n            predictions, (rmse, mae, r2, var) = evaluate(model, test, verbose=1)\n\n            signature = infer_signature(\n                train.select([\"_user_id\", \"_item_id\"]),\n                predictions.select([\"_user_id\", \"_item_id\", \"prediction\"]),\n            )\n            print(\"log model:\")\n            mlflow.spark.log_model(\n                model,\n                f\"{EXPERIMENT_NAME}-alsmodel\",\n                signature=signature,\n                registered_model_name=f\"{EXPERIMENT_NAME...",
          "displayName": "count at ALS.scala:1075",
          "jobGroup": "32",
          "jobId": 52,
          "killedTasksSummary": {},
          "name": "count at ALS.scala:1075",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 30,
          "numCompletedStages": 3,
          "numCompletedTasks": 30,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 6,
          "numSkippedTasks": 217,
          "numTasks": 247,
          "rowCount": 460,
          "stageIds": [
           117,
           114,
           118,
           115,
           111,
           112,
           116,
           113,
           110
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:37:40.910GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:37:40.844GMT",
          "dataRead": 45002,
          "dataWritten": 9386,
          "description": "Job group for statement 32:\nfrom mlflow.models.signature import infer_signature\n\nwith mlflow.start_run(run_name=\"als\"):\n    # Train models\n    models = tuner.fit(train)\n    best_metrics = {\"RMSE\": 10e6, \"MAE\": 10e6, \"R2\": 0, \"Explained variance\": 0}\n    best_index = 0\n    # Evaluate models\n    # Log model, metrics and parameters\n    for idx, model in enumerate(models.subModels):\n        with mlflow.start_run(nested=True, run_name=f\"als_{idx}\") as run:\n            print(\"\nEvaluating on testing data:\")\n            print(f\"subModel No. {idx + 1}\")\n            predictions, (rmse, mae, r2, var) = evaluate(model, test, verbose=1)\n\n            signature = infer_signature(\n                train.select([\"_user_id\", \"_item_id\"]),\n                predictions.select([\"_user_id\", \"_item_id\", \"prediction\"]),\n            )\n            print(\"log model:\")\n            mlflow.spark.log_model(\n                model,\n                f\"{EXPERIMENT_NAME}-alsmodel\",\n                signature=signature,\n                registered_model_name=f\"{EXPERIMENT_NAME...",
          "displayName": "count at ALS.scala:980",
          "jobGroup": "32",
          "jobId": 51,
          "killedTasksSummary": {},
          "name": "count at ALS.scala:980",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 11,
          "numCompletedStages": 2,
          "numCompletedTasks": 11,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 4,
          "numSkippedTasks": 215,
          "numTasks": 226,
          "rowCount": 300,
          "stageIds": [
           107,
           104,
           108,
           105,
           109,
           106
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:37:38.980GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:37:38.735GMT",
          "dataRead": 427173,
          "dataWritten": 17077,
          "description": "Job group for statement 32:\nfrom mlflow.models.signature import infer_signature\n\nwith mlflow.start_run(run_name=\"als\"):\n    # Train models\n    models = tuner.fit(train)\n    best_metrics = {\"RMSE\": 10e6, \"MAE\": 10e6, \"R2\": 0, \"Explained variance\": 0}\n    best_index = 0\n    # Evaluate models\n    # Log model, metrics and parameters\n    for idx, model in enumerate(models.subModels):\n        with mlflow.start_run(nested=True, run_name=f\"als_{idx}\") as run:\n            print(\"\nEvaluating on testing data:\")\n            print(f\"subModel No. {idx + 1}\")\n            predictions, (rmse, mae, r2, var) = evaluate(model, test, verbose=1)\n\n            signature = infer_signature(\n                train.select([\"_user_id\", \"_item_id\"]),\n                predictions.select([\"_user_id\", \"_item_id\", \"prediction\"]),\n            )\n            print(\"log model:\")\n            mlflow.spark.log_model(\n                model,\n                f\"{EXPERIMENT_NAME}-alsmodel\",\n                signature=signature,\n                registered_model_name=f\"{EXPERIMENT_NAME...",
          "displayName": "count at ALS.scala:973",
          "jobGroup": "32",
          "jobId": 50,
          "killedTasksSummary": {},
          "name": "count at ALS.scala:973",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 12,
          "numCompletedStages": 3,
          "numCompletedTasks": 12,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 3,
          "numSkippedTasks": 214,
          "numTasks": 226,
          "rowCount": 401,
          "stageIds": [
           102,
           99,
           103,
           100,
           101,
           98
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:37:31.275GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:37:30.797GMT",
          "dataRead": 220031501,
          "dataWritten": 124527155,
          "description": "Job group for statement 32:\nfrom mlflow.models.signature import infer_signature\n\nwith mlflow.start_run(run_name=\"als\"):\n    # Train models\n    models = tuner.fit(train)\n    best_metrics = {\"RMSE\": 10e6, \"MAE\": 10e6, \"R2\": 0, \"Explained variance\": 0}\n    best_index = 0\n    # Evaluate models\n    # Log model, metrics and parameters\n    for idx, model in enumerate(models.subModels):\n        with mlflow.start_run(nested=True, run_name=f\"als_{idx}\") as run:\n            print(\"\nEvaluating on testing data:\")\n            print(f\"subModel No. {idx + 1}\")\n            predictions, (rmse, mae, r2, var) = evaluate(model, test, verbose=1)\n\n            signature = infer_signature(\n                train.select([\"_user_id\", \"_item_id\"]),\n                predictions.select([\"_user_id\", \"_item_id\", \"prediction\"]),\n            )\n            print(\"log model:\")\n            mlflow.spark.log_model(\n                model,\n                f\"{EXPERIMENT_NAME}-alsmodel\",\n                signature=signature,\n                registered_model_name=f\"{EXPERIMENT_NAME...",
          "displayName": "isEmpty at ALS.scala:960",
          "jobGroup": "32",
          "jobId": 49,
          "killedTasksSummary": {},
          "name": "isEmpty at ALS.scala:960",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 201,
          "numCompletedStages": 2,
          "numCompletedTasks": 201,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 2,
          "numSkippedTasks": 14,
          "numTasks": 215,
          "rowCount": 1953090,
          "stageIds": [
           96,
           97,
           94,
           95
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:37:13.900GMT",
          "usageDescription": ""
         }
        ],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 18,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 32
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 32, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Evaluating on testing data:\n",
      "subModel No. 1\n",
      "RMSE score = 0.8566210059249003\n",
      "MAE score = 0.1821588006886569\n",
      "R2 score = 0.7106032110200756\n",
      "Explained variance = 3.6549009650456075\n",
      "log model:\n",
      "log metrics:\n",
      "log parameters:\n",
      "\n",
      "Evaluating on testing data:\n",
      "subModel No. 2\n",
      "RMSE score = 0.7864884691975794\n",
      "MAE score = 0.19574565407764388\n",
      "R2 score = 0.6559050461291518\n",
      "Explained variance = 3.5250225745824895\n",
      "log model:\n",
      "log metrics:\n",
      "log parameters:\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/trusted-service-user/cluster-env/trident_env/lib/python3.10/site-packages/mlflow/models/signature.py:130: UserWarning: Hint: Inferred schema contains integer column(s). Integer columns in Python cannot represent missing values. If your input data contains missing values at inference time, it will be encoded as floats and will cause a schema enforcement error. The best way to avoid this problem is to infer the model schema based on a realistic data sample (training dataset) that includes missing values. Alternatively, you can declare integer columns as doubles (float64) whenever these columns may have missing values. See `Handling Integers With Missing Values <https://www.mlflow.org/docs/latest/models.html#handling-integers-with-missing-values>`_ for more details.\n",
      "  inputs = _infer_schema(model_input)\n",
      "/home/trusted-service-user/cluster-env/trident_env/lib/python3.10/site-packages/mlflow/models/signature.py:131: UserWarning: Hint: Inferred schema contains integer column(s). Integer columns in Python cannot represent missing values. If your input data contains missing values at inference time, it will be encoded as floats and will cause a schema enforcement error. The best way to avoid this problem is to infer the model schema based on a realistic data sample (training dataset) that includes missing values. Alternatively, you can declare integer columns as doubles (float64) whenever these columns may have missing values. See `Handling Integers With Missing Values <https://www.mlflow.org/docs/latest/models.html#handling-integers-with-missing-values>`_ for more details.\n",
      "  outputs = _infer_schema(model_output) if model_output is not None else None\n",
      "2023/09/11 14:39:32 WARNING mlflow.utils.environment: Encountered an unexpected error while inferring pip requirements (model URI: /tmp/tmp0ejzcrc0/model, flavor: spark), fall back to return ['pyspark==3.3.1']. Set logging level to DEBUG to see the full traceback.\n",
      "/home/trusted-service-user/cluster-env/trident_env/lib/python3.10/site-packages/_distutils_hack/__init__.py:33: UserWarning: Setuptools is replacing distutils.\n",
      "  warnings.warn(\"Setuptools is replacing distutils.\")\n",
      "Successfully registered model 'aisample-recommendation-alsmodel'.\n",
      "2023/09/11 14:39:47 INFO mlflow.tracking._model_registry.client: Waiting up to 300 seconds for model version to finish creation.                     Model name: aisample-recommendation-alsmodel, version 4\n",
      "Created version '4' of model 'aisample-recommendation-alsmodel'.\n",
      "2023/09/11 14:40:46 WARNING mlflow.utils.environment: Encountered an unexpected error while inferring pip requirements (model URI: /tmp/tmplhzuq0ib/model, flavor: spark), fall back to return ['pyspark==3.3.1']. Set logging level to DEBUG to see the full traceback.\n",
      "Successfully registered model 'aisample-recommendation-alsmodel'.\n",
      "2023/09/11 14:40:59 INFO mlflow.tracking._model_registry.client: Waiting up to 300 seconds for model version to finish creation.                     Model name: aisample-recommendation-alsmodel, version 5\n",
      "Created version '5' of model 'aisample-recommendation-alsmodel'.\n",
      "2023/09/11 14:41:14 WARNING mlflow.utils.environment: Encountered an unexpected error while inferring pip requirements (model URI: /tmp/tmpg4wgzo9h/model, flavor: spark), fall back to return ['pyspark==3.3.1']. Set logging level to DEBUG to see the full traceback.\n",
      "Successfully registered model 'aisample-recommendation-alsmodel'.\n",
      "2023/09/11 14:41:27 INFO mlflow.tracking._model_registry.client: Waiting up to 300 seconds for model version to finish creation.                     Model name: aisample-recommendation-alsmodel, version 6\n",
      "Created version '6' of model 'aisample-recommendation-alsmodel'.\n"
     ]
    }
   ],
   "source": [
    "from mlflow.models.signature import infer_signature\n",
    "\n",
    "with mlflow.start_run(run_name=\"als\"):\n",
    "    # Train models\n",
    "    models = tuner.fit(train)\n",
    "    best_metrics = {\"RMSE\": 10e6, \"MAE\": 10e6, \"R2\": 0, \"Explained variance\": 0}\n",
    "    best_index = 0\n",
    "    # Evaluate models\n",
    "    # Log model, metrics and parameters\n",
    "    for idx, model in enumerate(models.subModels):\n",
    "        with mlflow.start_run(nested=True, run_name=f\"als_{idx}\") as run:\n",
    "            print(\"\\nEvaluating on testing data:\")\n",
    "            print(f\"subModel No. {idx + 1}\")\n",
    "            predictions, (rmse, mae, r2, var) = evaluate(model, test, verbose=1)\n",
    "\n",
    "            signature = infer_signature(\n",
    "                train.select([\"_user_id\", \"_item_id\"]),\n",
    "                predictions.select([\"_user_id\", \"_item_id\", \"prediction\"]),\n",
    "            )\n",
    "            print(\"log model:\")\n",
    "            mlflow.spark.log_model(\n",
    "                model,\n",
    "                f\"{EXPERIMENT_NAME}-alsmodel\",\n",
    "                signature=signature,\n",
    "                registered_model_name=f\"{EXPERIMENT_NAME}-alsmodel\",\n",
    "                dfs_tmpdir=\"Files/spark\",\n",
    "            )\n",
    "            print(\"log metrics:\")\n",
    "            current_metric = {\n",
    "                \"RMSE\": rmse,\n",
    "                \"MAE\": mae,\n",
    "                \"R2\": r2,\n",
    "                \"Explained variance\": var,\n",
    "            }\n",
    "            mlflow.log_metrics(current_metric)\n",
    "            if rmse < best_metrics[\"RMSE\"]:\n",
    "                best_metrics = current_metric\n",
    "                best_index = idx\n",
    "\n",
    "            print(\"log parameters:\")\n",
    "            mlflow.log_params(\n",
    "                {\n",
    "                    \"subModel_idx\": idx,\n",
    "                    \"num_epochs\": num_epochs,\n",
    "                    \"rank_size_list\": rank_size_list,\n",
    "                    \"reg_param_list\": reg_param_list,\n",
    "                    \"model_tuning_method\": model_tuning_method,\n",
    "                    \"DATA_FOLDER\": DATA_FOLDER,\n",
    "                }\n",
    "            )\n",
    "    # Log best model and related metrics and parameters to the parent run\n",
    "    mlflow.spark.log_model(\n",
    "        models.subModels[best_index],\n",
    "        f\"{EXPERIMENT_NAME}-alsmodel\",\n",
    "        signature=signature,\n",
    "        registered_model_name=f\"{EXPERIMENT_NAME}-alsmodel\",\n",
    "        dfs_tmpdir=\"Files/spark\",\n",
    "    )\n",
    "    mlflow.log_metrics(best_metrics)\n",
    "    mlflow.log_params(\n",
    "        {\n",
    "            \"subModel_idx\": idx,\n",
    "            \"num_epochs\": num_epochs,\n",
    "            \"rank_size_list\": rank_size_list,\n",
    "            \"reg_param_list\": reg_param_list,\n",
    "            \"model_tuning_method\": model_tuning_method,\n",
    "            \"DATA_FOLDER\": DATA_FOLDER,\n",
    "        }\n",
    "    )"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0c34b00f-2f22-4aaf-aaec-1397c1e529b1",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "To view the logged information for the training run, select the experiment named `aisample-recommendation` from your workspace. If you changed the experiment name, select the experiment with the name you specified. The logged information appears similar to the following images. Any experiment with its respective name is logged and you can track its parameters and performance metrics: \n",
    "\n",
    "<img style=\"float: left;\" src=\"https://synapseaisolutionsa.blob.core.windows.net/public/Book-Recommendation-Dataset/book-mlflow.png\"  width=\"45%\" height=\"10%\">\n",
    "<img style=\"float: left;\" src=\"https://synapseaisolutionsa.blob.core.windows.net/public/Book-Recommendation-Dataset/book-logging.png\"  width=\"45%\" height=\"10%\"> \n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e674d0ce-4bac-4a83-825b-a0e892e3608f",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "## Step 4: Load the final model for scoring and make predictions"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5aad60fc-7445-4253-bb45-898dc4f12e79",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "### Offline recommendation\n",
    "Once done with the training and having selected the best model, load the final model for scoring and generate predictions to recommend the top 10 books for each user. Also present the recommendations in an interpretable format."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "234e29d5-bfd2-4a25-a76e-7e033cafb9d7",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:41:41.1424339Z",
       "execution_start_time": "2023-09-11T14:41:30.5248694Z",
       "livy_statement_state": "available",
       "parent_msg_id": "57b258f2-7c43-4b67-a0ee-cc1de461ea72",
       "queued_time": "2023-09-11T14:33:16.492474Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [
         {
          "completionTime": "2023-09-11T14:41:39.135GMT",
          "dataRead": 0,
          "dataWritten": 0,
          "description": "Job group for statement 33:\n# Load the best model\n# Note that mlflow uses the PipelineModel to wrap the original model, thus we extract the original ALSModel from the stages\nmodel_uri = f\"models:/{EXPERIMENT_NAME}-alsmodel/1\"\nloaded_model = mlflow.spark.load_model(model_uri, dfs_tmpdir=\"Files/spark\").stages[-1]",
          "displayName": "load at ALS.scala:559",
          "jobGroup": "33",
          "jobId": 141,
          "killedTasksSummary": {},
          "name": "load at ALS.scala:559",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 1,
          "rowCount": 0,
          "stageIds": [
           496
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:41:38.271GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:41:38.066GMT",
          "dataRead": 0,
          "dataWritten": 0,
          "description": "Job group for statement 33:\n# Load the best model\n# Note that mlflow uses the PipelineModel to wrap the original model, thus we extract the original ALSModel from the stages\nmodel_uri = f\"models:/{EXPERIMENT_NAME}-alsmodel/1\"\nloaded_model = mlflow.spark.load_model(model_uri, dfs_tmpdir=\"Files/spark\").stages[-1]",
          "displayName": "load at ALS.scala:557",
          "jobGroup": "33",
          "jobId": 140,
          "killedTasksSummary": {},
          "name": "load at ALS.scala:557",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 1,
          "rowCount": 0,
          "stageIds": [
           495
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:41:37.215GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:41:36.973GMT",
          "dataRead": 378,
          "dataWritten": 0,
          "description": "Job group for statement 33:\n# Load the best model\n# Note that mlflow uses the PipelineModel to wrap the original model, thus we extract the original ALSModel from the stages\nmodel_uri = f\"models:/{EXPERIMENT_NAME}-alsmodel/1\"\nloaded_model = mlflow.spark.load_model(model_uri, dfs_tmpdir=\"Files/spark\").stages[-1]",
          "displayName": "first at ReadWrite.scala:587",
          "jobGroup": "33",
          "jobId": 139,
          "killedTasksSummary": {},
          "name": "first at ReadWrite.scala:587",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 1,
          "rowCount": 1,
          "stageIds": [
           494
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:41:36.912GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:41:36.766GMT",
          "dataRead": 378,
          "dataWritten": 0,
          "description": "Job group for statement 33:\n# Load the best model\n# Note that mlflow uses the PipelineModel to wrap the original model, thus we extract the original ALSModel from the stages\nmodel_uri = f\"models:/{EXPERIMENT_NAME}-alsmodel/1\"\nloaded_model = mlflow.spark.load_model(model_uri, dfs_tmpdir=\"Files/spark\").stages[-1]",
          "displayName": "first at ReadWrite.scala:587",
          "jobGroup": "33",
          "jobId": 138,
          "killedTasksSummary": {},
          "name": "first at ReadWrite.scala:587",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 1,
          "rowCount": 1,
          "stageIds": [
           493
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:41:36.691GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:41:36.588GMT",
          "dataRead": 210,
          "dataWritten": 0,
          "description": "Job group for statement 33:\n# Load the best model\n# Note that mlflow uses the PipelineModel to wrap the original model, thus we extract the original ALSModel from the stages\nmodel_uri = f\"models:/{EXPERIMENT_NAME}-alsmodel/1\"\nloaded_model = mlflow.spark.load_model(model_uri, dfs_tmpdir=\"Files/spark\").stages[-1]",
          "displayName": "first at ReadWrite.scala:587",
          "jobGroup": "33",
          "jobId": 137,
          "killedTasksSummary": {},
          "name": "first at ReadWrite.scala:587",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 1,
          "rowCount": 1,
          "stageIds": [
           492
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:41:36.132GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:41:35.934GMT",
          "dataRead": 210,
          "dataWritten": 0,
          "description": "Job group for statement 33:\n# Load the best model\n# Note that mlflow uses the PipelineModel to wrap the original model, thus we extract the original ALSModel from the stages\nmodel_uri = f\"models:/{EXPERIMENT_NAME}-alsmodel/1\"\nloaded_model = mlflow.spark.load_model(model_uri, dfs_tmpdir=\"Files/spark\").stages[-1]",
          "displayName": "runJob at PythonRDD.scala:172",
          "jobGroup": "33",
          "jobId": 136,
          "killedTasksSummary": {},
          "name": "runJob at PythonRDD.scala:172",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 1,
          "rowCount": 1,
          "stageIds": [
           491
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:41:31.715GMT",
          "usageDescription": ""
         }
        ],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 6,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 33
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 33, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023/09/11 14:41:30 INFO mlflow.spark: 'models:/aisample-recommendation-alsmodel/1' resolved as 'abfss://33953aa3-b8e7-44f4-a8aa-0c2588e6ce74@onelakemsit.pbidedicated.windows.net/fc42012a-38f8-4b16-8396-35751718fbac/3a8e0332-bc1a-4fbd-9630-e48474d81f29/artifacts'\n",
      "2023/09/11 14:41:31 INFO mlflow.spark: File 'abfss://33953aa3-b8e7-44f4-a8aa-0c2588e6ce74@onelakemsit.pbidedicated.windows.net/fc42012a-38f8-4b16-8396-35751718fbac/3a8e0332-bc1a-4fbd-9630-e48474d81f29/artifacts/sparkml' is already on DFS, copy is not necessary.\n"
     ]
    }
   ],
   "source": [
    "# Load the best model\n",
    "# MLflow uses the PipelineModel to wrap the original model to extract the original ALSModel from the stages\n",
    "model_uri = f\"models:/{EXPERIMENT_NAME}-alsmodel/1\"\n",
    "loaded_model = mlflow.spark.load_model(model_uri, dfs_tmpdir=\"Files/spark\").stages[-1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d54b4290-5c72-492e-9c0b-4dc488bc068a",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:41:42.0384296Z",
       "execution_start_time": "2023-09-11T14:41:41.7040789Z",
       "livy_statement_state": "available",
       "parent_msg_id": "d09d8281-d473-4861-acf8-3bf1bb0046b5",
       "queued_time": "2023-09-11T14:33:17.8128645Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 0,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 34
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 34, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Generate top 10 book recommendations for each user\n",
    "userRecs = loaded_model.recommendForAllUsers(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e8e75b41-0db0-4682-873d-f48138b74a5b",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:41:49.21306Z",
       "execution_start_time": "2023-09-11T14:41:42.4858974Z",
       "livy_statement_state": "available",
       "parent_msg_id": "06fd3d40-39b1-4004-bfeb-18352fcfc5b3",
       "queued_time": "2023-09-11T14:33:18.8512926Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [
         {
          "completionTime": "2023-09-11T14:41:47.761GMT",
          "dataRead": 0,
          "dataWritten": 0,
          "description": "Job group for statement 35:\n# Represent the recommendations in an interpretable format\nuserRecs = (\n    userRecs.withColumn(\"rec_exp\", F.explode(\"recommendations\"))\n    .select(\"_user_id\", F.col(\"rec_exp._item_id\"), F.col(\"rec_exp.rating\"))\n    .join(df_items.select([\"_item_id\", \"Book-Title\"]), on=\"_item_id\")\n)\nuserRecs.limit(10).show()",
          "displayName": "showString at NativeMethodAccessorImpl.java:0",
          "jobGroup": "35",
          "jobId": 146,
          "killedTasksSummary": {},
          "name": "showString at NativeMethodAccessorImpl.java:0",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 2,
          "numSkippedTasks": 11,
          "numTasks": 12,
          "rowCount": 0,
          "stageIds": [
           502,
           503,
           504
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:41:47.738GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:41:47.719GMT",
          "dataRead": 0,
          "dataWritten": 0,
          "description": "Job group for statement 35:\n# Represent the recommendations in an interpretable format\nuserRecs = (\n    userRecs.withColumn(\"rec_exp\", F.explode(\"recommendations\"))\n    .select(\"_user_id\", F.col(\"rec_exp._item_id\"), F.col(\"rec_exp.rating\"))\n    .join(df_items.select([\"_item_id\", \"Book-Title\"]), on=\"_item_id\")\n)\nuserRecs.limit(10).show()",
          "displayName": "showString at NativeMethodAccessorImpl.java:0",
          "jobGroup": "35",
          "jobId": 145,
          "killedTasksSummary": {},
          "name": "showString at NativeMethodAccessorImpl.java:0",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 1,
          "numSkippedTasks": 10,
          "numTasks": 11,
          "rowCount": 0,
          "stageIds": [
           500,
           501
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:41:47.560GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:41:47.535GMT",
          "dataRead": 201921,
          "dataWritten": 655207,
          "description": "Job group for statement 35:\n# Represent the recommendations in an interpretable format\nuserRecs = (\n    userRecs.withColumn(\"rec_exp\", F.explode(\"recommendations\"))\n    .select(\"_user_id\", F.col(\"rec_exp._item_id\"), F.col(\"rec_exp.rating\"))\n    .join(df_items.select([\"_item_id\", \"Book-Title\"]), on=\"_item_id\")\n)\nuserRecs.limit(10).show()",
          "displayName": "showString at NativeMethodAccessorImpl.java:0",
          "jobGroup": "35",
          "jobId": 144,
          "killedTasksSummary": {},
          "name": "showString at NativeMethodAccessorImpl.java:0",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 10,
          "numCompletedStages": 1,
          "numCompletedTasks": 10,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 10,
          "rowCount": 3771,
          "stageIds": [
           499
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:41:46.115GMT",
          "usageDescription": ""
         }
        ],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 3,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 35
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 35, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+--------+--------+---------+--------------------+\n",
      "|_item_id|_user_id|   rating|          Book-Title|\n",
      "+--------+--------+---------+--------------------+\n",
      "|   71056|       7|10.227177|   The Da Vinci Code|\n",
      "|  170437|       7|8.2797165|The Five People Y...|\n",
      "|   44865|       7|7.9993467|Lasher: Lives of ...|\n",
      "|  125415|       7|6.9831495|Harry Potter and ...|\n",
      "|   86727|       7| 6.354976|     No Greater Love|\n",
      "|    7481|       7|6.0774207|The Poisonwood Bi...|\n",
      "|  169605|       7| 6.069703|     Blood and Money|\n",
      "|  236327|       7|6.0508156|  Concept of Justice|\n",
      "|   70557|       7|5.9909506|Tuesdays with Mor...|\n",
      "|   86654|       7|5.8801236|      A Time to Kill|\n",
      "+--------+--------+---------+--------------------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Represent the recommendations in an interpretable format\n",
    "userRecs = (\n",
    "    userRecs.withColumn(\"rec_exp\", F.explode(\"recommendations\"))\n",
    "    .select(\"_user_id\", F.col(\"rec_exp._item_id\"), F.col(\"rec_exp.rating\"))\n",
    "    .join(df_items.select([\"_item_id\", \"Book-Title\"]), on=\"_item_id\")\n",
    ")\n",
    "userRecs.limit(10).show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ae886a11-4390-4cd9-8559-18296cc22829",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "You have received the top 10 book recommendations for each user."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5e8dfafe-c21b-4277-bed4-cee5e827b006",
   "metadata": {
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "source": [
    "#### Save the predictions to the lakehouse\n",
    "\n",
    "Finally, write the recommendation generated by the best ALS model back to the lakehouse."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "0e3c64a3-8ea2-4ffe-873b-386b63f1050c",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:41:56.4986393Z",
       "execution_start_time": "2023-09-11T14:41:49.6550868Z",
       "livy_statement_state": "available",
       "parent_msg_id": "e0ac0653-0ddc-4ea5-b0c0-797db749a762",
       "queued_time": "2023-09-11T14:33:19.7447937Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [
         {
          "completionTime": "2023-09-11T14:41:55.025GMT",
          "dataRead": 0,
          "dataWritten": 0,
          "description": "Delta: Job group for statement 36:\n# Code to save the userRecs into lakehouse\nuserRecs.write.format(\"delta\").mode(\"overwrite\").save(\n    f\"{DATA_FOLDER}/predictions/userRecs\"\n): Compute snapshot for version: 0",
          "displayName": "toString at String.java:2994",
          "jobGroup": "36",
          "jobId": 155,
          "killedTasksSummary": {},
          "name": "toString at String.java:2994",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 2,
          "numSkippedTasks": 51,
          "numTasks": 52,
          "rowCount": 0,
          "stageIds": [
           517,
           518,
           516
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:41:54.961GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:41:54.942GMT",
          "dataRead": 0,
          "dataWritten": 0,
          "description": "Delta: Job group for statement 36:\n# Code to save the userRecs into lakehouse\nuserRecs.write.format(\"delta\").mode(\"overwrite\").save(\n    f\"{DATA_FOLDER}/predictions/userRecs\"\n): Compute snapshot for version: 0",
          "displayName": "toString at String.java:2994",
          "jobGroup": "36",
          "jobId": 154,
          "killedTasksSummary": {},
          "name": "toString at String.java:2994",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 50,
          "numCompletedStages": 1,
          "numCompletedTasks": 50,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 1,
          "numSkippedTasks": 1,
          "numTasks": 51,
          "rowCount": 0,
          "stageIds": [
           514,
           515
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:41:53.972GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:41:53.663GMT",
          "dataRead": 0,
          "dataWritten": 0,
          "description": "Delta: Job group for statement 36:\n# Code to save the userRecs into lakehouse\nuserRecs.write.format(\"delta\").mode(\"overwrite\").save(\n    f\"{DATA_FOLDER}/predictions/userRecs\"\n): Compute snapshot for version: 0",
          "displayName": "toString at String.java:2994",
          "jobGroup": "36",
          "jobId": 153,
          "killedTasksSummary": {},
          "name": "toString at String.java:2994",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 1,
          "rowCount": 0,
          "stageIds": [
           513
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:41:53.238GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:41:52.678GMT",
          "dataRead": 0,
          "dataWritten": 0,
          "description": "Job group for statement 36:\n# Code to save the userRecs into lakehouse\nuserRecs.write.format(\"delta\").mode(\"overwrite\").save(\n    f\"{DATA_FOLDER}/predictions/userRecs\"\n)",
          "displayName": "Job group for statement 36:\n# Code to save the userRecs into lakehouse\nuserRecs.write.format(\"delta\").mode(\"overwrite\").save(\n    f\"{DATA_FOLDER}/predictions/userRecs\"\n)",
          "jobGroup": "36",
          "jobId": 152,
          "killedTasksSummary": {},
          "name": "",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 0,
          "numCompletedStages": 0,
          "numCompletedTasks": 0,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 0,
          "rowCount": 0,
          "stageIds": [],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:41:52.678GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:41:52.540GMT",
          "dataRead": 0,
          "dataWritten": 0,
          "description": "Job group for statement 36:\n# Code to save the userRecs into lakehouse\nuserRecs.write.format(\"delta\").mode(\"overwrite\").save(\n    f\"{DATA_FOLDER}/predictions/userRecs\"\n)",
          "displayName": "save at NativeMethodAccessorImpl.java:0",
          "jobGroup": "36",
          "jobId": 151,
          "killedTasksSummary": {},
          "name": "save at NativeMethodAccessorImpl.java:0",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 2,
          "numSkippedTasks": 11,
          "numTasks": 12,
          "rowCount": 0,
          "stageIds": [
           510,
           511,
           512
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:41:51.834GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:41:51.796GMT",
          "dataRead": 0,
          "dataWritten": 0,
          "description": "Job group for statement 36:\n# Code to save the userRecs into lakehouse\nuserRecs.write.format(\"delta\").mode(\"overwrite\").save(\n    f\"{DATA_FOLDER}/predictions/userRecs\"\n)",
          "displayName": "save at NativeMethodAccessorImpl.java:0",
          "jobGroup": "36",
          "jobId": 150,
          "killedTasksSummary": {},
          "name": "save at NativeMethodAccessorImpl.java:0",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 1,
          "numCompletedStages": 1,
          "numCompletedTasks": 1,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 1,
          "numSkippedTasks": 10,
          "numTasks": 11,
          "rowCount": 0,
          "stageIds": [
           509,
           508
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:41:51.554GMT",
          "usageDescription": ""
         },
         {
          "completionTime": "2023-09-11T14:41:51.530GMT",
          "dataRead": 0,
          "dataWritten": 0,
          "description": "Job group for statement 36:\n# Code to save the userRecs into lakehouse\nuserRecs.write.format(\"delta\").mode(\"overwrite\").save(\n    f\"{DATA_FOLDER}/predictions/userRecs\"\n)",
          "displayName": "save at NativeMethodAccessorImpl.java:0",
          "jobGroup": "36",
          "jobId": 149,
          "killedTasksSummary": {},
          "name": "save at NativeMethodAccessorImpl.java:0",
          "numActiveStages": 0,
          "numActiveTasks": 0,
          "numCompletedIndices": 10,
          "numCompletedStages": 1,
          "numCompletedTasks": 10,
          "numFailedStages": 0,
          "numFailedTasks": 0,
          "numKilledTasks": 0,
          "numSkippedStages": 0,
          "numSkippedTasks": 0,
          "numTasks": 10,
          "rowCount": 0,
          "stageIds": [
           507
          ],
          "status": "SUCCEEDED",
          "submissionTime": "2023-09-11T14:41:50.819GMT",
          "usageDescription": ""
         }
        ],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 7,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 36
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 36, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Code to save the userRecs into lakehouse\n",
    "userRecs.write.format(\"delta\").mode(\"overwrite\").save(\n",
    "    f\"{DATA_FOLDER}/predictions/userRecs\"\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "36469bab-7db2-4b84-bf35-d320ba48ae7b",
   "metadata": {
    "jupyter": {
     "outputs_hidden": false,
     "source_hidden": false
    },
    "nteract": {
     "transient": {
      "deleting": false
     }
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.livy.statement-meta+json": {
       "execution_finish_time": "2023-09-11T14:41:57.2634658Z",
       "execution_start_time": "2023-09-11T14:41:56.9267795Z",
       "livy_statement_state": "available",
       "parent_msg_id": "2198affd-56de-4ef8-9960-9b52203aad67",
       "queued_time": "2023-09-11T14:33:20.410754Z",
       "session_id": "db09740a-036f-4e14-9afc-ee8d112e8535",
       "session_start_time": null,
       "spark_jobs": {
        "jobs": [],
        "limit": 20,
        "numbers": {
         "FAILED": 0,
         "RUNNING": 0,
         "SUCCEEDED": 0,
         "UNKNOWN": 0
        },
        "rule": "ALL_DESC"
       },
       "spark_pool": null,
       "state": "finished",
       "statement_id": 37
      },
      "text/plain": [
       "StatementMeta(, db09740a-036f-4e14-9afc-ee8d112e8535, 37, Finished, Available)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Full run cost 522 seconds.\n"
     ]
    }
   ],
   "source": [
    "# Determine the entire runtime\n",
    "print(f\"Full run cost {int(time.time() - ts)} seconds.\")"
   ]
  }
 ],
 "metadata": {
  "kernel_info": {
   "name": "synapse_pyspark"
  },
  "kernelspec": {
   "display_name": "Synapse PySpark",
   "language": "Python",
   "name": "synapse_pyspark"
  },
  "language_info": {
   "name": "python"
  },
  "notebook_environment": {},
  "nteract": {
   "version": "nteract-front-end@1.0.0"
  },
  "save_output": true,
  "spark_compute": {
   "compute_id": "/trident/default",
   "session_options": {
    "conf": {},
    "enableDebugMode": false
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}