notebook/4b_Modeling_Tuning_xgboost.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 141,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load the \"autoreload\" extension so that code can change\n",
    "# Always reload modules so that as soon as code changes in src, it gets automatically reloaded without kernel relaunch\n",
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "import os\n",
    "import sys\n",
    "sys.path.append('../') \n",
    "\n",
    "from src.utils import datacollector\n",
    "from src.utils import constants as cst\n",
    "\n",
    "# Visualization\n",
    "from src.visualization import visualization as viz\n",
    "\n",
    "# Models\n",
    "from src.modeling import modeling\n",
    "import xgboost as xgb\n",
    "from xgboost import plot_importance\n",
    "\n",
    "from sklearn.metrics import mean_squared_error"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# REMINDER - OBJECTIVE\n",
    "In the previous [notebook](4_Modeling.ipynb), XGBoost model has been tried with few parameters tuned through a GridSearch. It gave us the best results but we have seen that there was high overfitting.  \n",
    "Here the goal is to **keep moving forward on parameters tuning and try to reduce overfitting** with some parameters made for that in addition with the use of the earlystopping functionnality.  \n",
    "We will **also use the validation dataset** as evaluation one so that we have train, validation for metric evaluation and earlystopping and then the test dataset.\n",
    "\n",
    "# 1. Load datasets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = pd.read_csv(datacollector.get_data_file(cst.LST_X_TRAIN_FILE, True), sep=',', header=0)\n",
    "y_train = pd.read_csv(datacollector.get_data_file(cst.LST_Y_TRAIN_FILE, True), sep=',', header=None)\n",
    "X_val = pd.read_csv(datacollector.get_data_file(cst.LST_X_VAL_FILE, True), sep=',', header=0)\n",
    "y_val = pd.read_csv(datacollector.get_data_file(cst.LST_Y_VAL_FILE, True), sep=',', header=None)\n",
    "X_test = pd.read_csv(datacollector.get_data_file(cst.LST_X_TEST_FILE, True), sep=',', header=0)\n",
    "y_test = pd.read_csv(datacollector.get_data_file(cst.LST_Y_TEST_FILE, True), sep=',', header=None)\n",
    "\n",
    "y_train.columns = ['train_price']\n",
    "y_val.columns = ['val_price']\n",
    "y_test.columns = ['test_price']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X_train.shape is (52076, 153)\n",
      "y_train.shape is (52076, 1)\n",
      "X_val.shape is (5787, 153)\n",
      "y_val.shape is (5787, 1)\n",
      "X_test.shape is (6430, 153)\n",
      "y_test.shape is (6430, 1)\n"
     ]
    }
   ],
   "source": [
    "print(\"X_train.shape is {}\".format(X_train.shape))\n",
    "print(\"y_train.shape is {}\".format(y_train.shape))\n",
    "print(\"X_val.shape is {}\".format(X_val.shape))\n",
    "print(\"y_val.shape is {}\".format(y_val.shape))\n",
    "print(\"X_test.shape is {}\".format(X_test.shape))\n",
    "print(\"y_test.shape is {}\".format(y_test.shape))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "# 2. RandomizedSearchCV instead of classic GridSearchCV\n",
    "### 2.1. First run\n",
    "See documentation [here](https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.RandomizedSearchCV.html) for more details and examples about RandomizedSearchCV.  \n",
    "Explore values for `reg_alpha` parameter which is a regularization one."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Parameters of xgboost that may change\n",
    "param_grid = {\n",
    "    'n_estimators': [100, 200, 300], # Number of estimators (weak learners)\n",
    "    'min_child_weight': [4, 5, 6],\n",
    "    'learning_rate': [0.01, 0.03, 0.1],\n",
    "    'subsample': [i/10.0 for i in range(6, 10)],\n",
    "    'colsample_bytree': [i/10.0 for i in range(6, 10)], # subsample ratio of columns when constructing each tree (range is [0, 1])\n",
    "    'reg_alpha': [0.02, 0.03, 0.04, 0.1, 1], # L1 regularization term on weights. Increasing this value will make model more conservative\n",
    "    'max_depth': [5, 6, 7]  # Maximum depth of a tree. Increasing this value will make the model more complex and more likely to overfit\n",
    "}\n",
    "\n",
    "xgb_randomsearch = modeling.build_xgb_random_search(param_grid, num_iters=40)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 40 candidates, totalling 200 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed: 18.7min\n",
      "[Parallel(n_jobs=-1)]: Done 184 tasks      | elapsed: 67.4min\n",
      "[Parallel(n_jobs=-1)]: Done 200 out of 200 | elapsed: 72.8min finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "RMSE for tuned and optimized xgboost model on train: 90.87\n",
      "RMSE for tuned and optimized xgboost model on test: 102.38\n"
     ]
    }
   ],
   "source": [
    "y_xgbrs_run1 = modeling.fit_and_run_pipeline(xgb_randomsearch, 'tuned and optimized xgboost', X_train, y_train, X_test, y_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Observations:**\n",
    "* **No significant improvement for RMSE value on test dataset**\n",
    "* but we can see that there is less difference between train and test so I think **we managed to reduce** a little bit the **overfitting**.\n",
    "* It seems to be really difficult to reduce this RMSE value which remains high. Perhaps this is the **best we can achieve with this number of estimators** (our best one has 300 which is the highest value given in the parameters grid). **We should try to increase this value**.\n",
    "\n",
    "### 2.2. Feature importances"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 936x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "viz.plot_feature_importances(xgb_randomsearch.best_estimator_.feature_importances_, X_test.columns.tolist())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Observations**:\n",
    "* We find again the `cancellation_policy_super_strict_60` that the basic DecisionTree found as well but now `bathrooms` is as important.\n",
    "* Less `neighbourhoods` in this TOP 25 (Vaugirard, Montmartre) and more amenities\n",
    "* `Safety card` is still here (3rd)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###  2.3. Results comparison table & error analysis\n",
    "To have a better intuition on how 'accurate' our models were, I will concatenate all predictions for our models and compare that to the ground truth. This results file will be saved so that it can be loaded later for further investigations."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "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>y_true</th>\n",
       "      <th>y_xgbrs_run1</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>59.0</td>\n",
       "      <td>63.762135</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>90.0</td>\n",
       "      <td>80.044769</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>46.0</td>\n",
       "      <td>55.033432</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>67.0</td>\n",
       "      <td>90.333168</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>59.0</td>\n",
       "      <td>84.442856</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>95.0</td>\n",
       "      <td>108.445274</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>90.0</td>\n",
       "      <td>78.571976</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>119.0</td>\n",
       "      <td>514.855530</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>90.0</td>\n",
       "      <td>130.405289</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>65.0</td>\n",
       "      <td>77.177063</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>85.0</td>\n",
       "      <td>104.448616</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>42.0</td>\n",
       "      <td>54.656857</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>50.0</td>\n",
       "      <td>63.454319</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>100.0</td>\n",
       "      <td>194.024139</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>100.0</td>\n",
       "      <td>85.301147</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>130.0</td>\n",
       "      <td>189.043747</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>65.0</td>\n",
       "      <td>67.120461</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>99.0</td>\n",
       "      <td>78.963585</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>250.0</td>\n",
       "      <td>160.048218</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>30.0</td>\n",
       "      <td>67.224190</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    y_true  y_xgbrs_run1\n",
       "0     59.0     63.762135\n",
       "1     90.0     80.044769\n",
       "2     46.0     55.033432\n",
       "3     67.0     90.333168\n",
       "4     59.0     84.442856\n",
       "5     95.0    108.445274\n",
       "6     90.0     78.571976\n",
       "7    119.0    514.855530\n",
       "8     90.0    130.405289\n",
       "9     65.0     77.177063\n",
       "10    85.0    104.448616\n",
       "11    42.0     54.656857\n",
       "12    50.0     63.454319\n",
       "13   100.0    194.024139\n",
       "14   100.0     85.301147\n",
       "15   130.0    189.043747\n",
       "16    65.0     67.120461\n",
       "17    99.0     78.963585\n",
       "18   250.0    160.048218\n",
       "19    30.0     67.224190"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results_df = pd.concat([y_test, pd.DataFrame(data=y_xgbrs_run1)],axis=1)\n",
    "results_df.columns = ['y_true', 'y_xgbrs_run1']\n",
    "results_df.head(20)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Observation:** even if value are different, at first glance, it seems that predictions are more or less the \"same\" than previous results.  \n",
    "Let's classify also to have a better understanding."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "results_df = modeling.classify_results(results_df, 'y_xgbrs_run1')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "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>y_true</th>\n",
       "      <th>y_xgbrs_run1</th>\n",
       "      <th>y_xgbrs_run1_perc_diff</th>\n",
       "      <th>y_xgbrs_run1_perc_diff_class</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>59.0</td>\n",
       "      <td>63.762135</td>\n",
       "      <td>8.07</td>\n",
       "      <td>3-good (5% &lt; x &lt; 10%)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>90.0</td>\n",
       "      <td>80.044769</td>\n",
       "      <td>11.06</td>\n",
       "      <td>4-acceptable (10% &lt; x &lt; 20%)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>46.0</td>\n",
       "      <td>55.033432</td>\n",
       "      <td>19.64</td>\n",
       "      <td>4-acceptable (10% &lt; x &lt; 20%)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>67.0</td>\n",
       "      <td>90.333168</td>\n",
       "      <td>34.83</td>\n",
       "      <td>5-bad (20% &lt; x &lt; 40%)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>59.0</td>\n",
       "      <td>84.442856</td>\n",
       "      <td>43.12</td>\n",
       "      <td>6-awful (&gt; 40%)</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   y_true  y_xgbrs_run1  y_xgbrs_run1_perc_diff  y_xgbrs_run1_perc_diff_class\n",
       "0    59.0     63.762135                    8.07         3-good (5% < x < 10%)\n",
       "1    90.0     80.044769                   11.06  4-acceptable (10% < x < 20%)\n",
       "2    46.0     55.033432                   19.64  4-acceptable (10% < x < 20%)\n",
       "3    67.0     90.333168                   34.83         5-bad (20% < x < 40%)\n",
       "4    59.0     84.442856                   43.12               6-awful (> 40%)"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1-very accurate (< 2%)           332\n",
       "2-pretty good (2% < x < 5%)      483\n",
       "3-good (5% < x < 10%)            837\n",
       "4-acceptable (10% < x < 20%)    1402\n",
       "5-bad (20% < x < 40%)           1854\n",
       "6-awful (> 40%)                 1522\n",
       "Name: y_xgbrs_run1_perc_diff_class, dtype: int64"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results_df['y_xgbrs_run1_perc_diff_class'].value_counts().sort_index()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "***Observation:*** classes 1 to 4 (included) counts for 47.5% of the dataset, we are still in the same proportions.  \n",
    "Let's try another approach.\n",
    "\n",
    "---\n",
    "# 3. Use python API (not the sklearn one)\n",
    "As said several times now, when using xgboost through its sklearn wrapper API, the xgboost's built-in `'earlystopping'` function cannot be used with GridSearchCV. There are few posts on the Internet talking about that (for example [here](https://discuss.xgboost.ai/t/how-to-do-early-stopping-with-scikit-learns-gridsearchcv/151) or [there](https://stackoverflow.com/questions/50824326/xgboost-with-gridsearchcv-scaling-pca-and-early-stopping-in-sklearn-pipeline)). One workaround is to use the xgboost built-in `cv` function and do the grid search by ourselves.\n",
    "\n",
    "The following section has been developped by following this [tutorial](https://blog.cambridgespark.com/hyperparameter-tuning-in-xgboost-4ff9100a3b2f) and also this [other one](https://www.analyticsvidhya.com/blog/2016/03/complete-guide-parameter-tuning-xgboost-with-codes-python/). Both articles describe the same approach, more or less, which is:\n",
    "1. Choose a relatively high learning rate and determine the optimum number of trees for this learning rate. XGBoost has a very useful function called as “cv” which performs cross-validation at each boosting iteration and thus returns the optimum number of trees required.\n",
    "2. Tune tree-specific parameters (`max_depth`, `min_child_weight`, `gamma`, `subsample`, `colsample_bytree`) for decided learning rate and number of trees.\n",
    "3. Tune regularization parameters (`lambda`, `alpha`) for xgboost which can help reduce model complexity and enhance performance.\n",
    "4. Lower the learning rate and decide the optimal parameters. \n",
    "\n",
    "(source: [here](https://www.analyticsvidhya.com/blog/2016/03/complete-guide-parameter-tuning-xgboost-with-codes-python/))\n",
    "\n",
    "### 3.1. Transform datasets into DMatrices"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# XGBoost has native handling of missing values and if nothing is given then np.nan is the value which is taken according to\n",
    "# the documentation here: https://xgboost.readthedocs.io/en/latest/python/python_api.html?highlight=xgbregressor#module-xgboost.core\n",
    "dtrain = xgb.DMatrix(X_train, label=y_train)\n",
    "dval = xgb.DMatrix(X_val, label=y_val)\n",
    "dtest = xgb.DMatrix(X_test, label=y_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.2. Set parameters grid to default values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 152,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0]\tValidation-rmse:162.661\n",
      "Will train until Validation-rmse hasn't improved in 30 rounds.\n",
      "[1]\tValidation-rmse:142.142\n",
      "[2]\tValidation-rmse:130.15\n",
      "[3]\tValidation-rmse:121.947\n",
      "[4]\tValidation-rmse:117.891\n",
      "[5]\tValidation-rmse:114.193\n",
      "[6]\tValidation-rmse:112.642\n",
      "[7]\tValidation-rmse:113.364\n",
      "[8]\tValidation-rmse:114.35\n",
      "[9]\tValidation-rmse:113.557\n",
      "[10]\tValidation-rmse:113.224\n",
      "[11]\tValidation-rmse:112.405\n",
      "[12]\tValidation-rmse:111.495\n",
      "[13]\tValidation-rmse:111.207\n",
      "[14]\tValidation-rmse:110.848\n",
      "[15]\tValidation-rmse:110.84\n",
      "[16]\tValidation-rmse:113.274\n",
      "[17]\tValidation-rmse:113.242\n",
      "[18]\tValidation-rmse:112.986\n",
      "[19]\tValidation-rmse:112.599\n",
      "[20]\tValidation-rmse:112.552\n",
      "[21]\tValidation-rmse:111.985\n",
      "[22]\tValidation-rmse:111.978\n",
      "[23]\tValidation-rmse:111.952\n",
      "[24]\tValidation-rmse:114.198\n",
      "[25]\tValidation-rmse:115.955\n",
      "[26]\tValidation-rmse:115.854\n",
      "[27]\tValidation-rmse:115.776\n",
      "[28]\tValidation-rmse:115.791\n",
      "[29]\tValidation-rmse:117.009\n",
      "[30]\tValidation-rmse:120.145\n",
      "[31]\tValidation-rmse:119.856\n",
      "[32]\tValidation-rmse:119.947\n",
      "[33]\tValidation-rmse:119.844\n",
      "[34]\tValidation-rmse:119.812\n",
      "[35]\tValidation-rmse:119.776\n",
      "[36]\tValidation-rmse:119.814\n",
      "[37]\tValidation-rmse:120.459\n",
      "[38]\tValidation-rmse:120.44\n",
      "[39]\tValidation-rmse:120.456\n",
      "[40]\tValidation-rmse:120.201\n",
      "[41]\tValidation-rmse:120.336\n",
      "[42]\tValidation-rmse:120.329\n",
      "[43]\tValidation-rmse:120.33\n",
      "[44]\tValidation-rmse:120.403\n",
      "[45]\tValidation-rmse:120.195\n",
      "Stopping. Best iteration:\n",
      "[15]\tValidation-rmse:110.84\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Parameters of xgboost that may change\n",
    "params = {   \n",
    "    # Defines the minimum sum of weights of all observations required in a child.\n",
    "    # Used to control over-fitting. Higher values prevent a model from learning relations which might be highly specific\n",
    "    # to the particular sample selected for a tree.\n",
    "    # [default=1]\n",
    "    'min_child_weight': 1,\n",
    "    \n",
    "    # Maximum depth of a tree. Increasing this value will make the model more complex and more likely to overfit\n",
    "    'max_depth': 6,\n",
    "    \n",
    "    # Gamma: A node is split only when the resulting split gives a positive reduction in the loss function.\n",
    "    # Gamma specifies the minimum loss reduction required to make a further partition on a leaf node of the tree.\n",
    "    # Makes the algorithm conservative.\n",
    "    'gamma': 0,\n",
    "    \n",
    "    # Learning rate\n",
    "    'eta': .3,\n",
    "    \n",
    "    # Denotes the fraction of observations to be randomly samples for each tree.\n",
    "    # Lower values make the algorithm more conservative and prevents overfitting but too small values might lead to \n",
    "    # under-fitting. Typical values: 0.5-1\n",
    "    # [default=1] means we use all rows\n",
    "    'subsample': 1,\n",
    "    \n",
    "    # Subsample ratio of columns when constructing each tree (range is [0, 1])\n",
    "    # [default=1] means we use all columns/features\n",
    "    'colsample_bytree': 1,\n",
    "    \n",
    "    # L1 regularization term on weights. Increasing this value will make model more conservative\n",
    "    # Can be used in case of very high dimensionality so that the algorithm runs faster when implemented\n",
    "    # [default=0]\n",
    "    'alpha': 0, \n",
    "    \n",
    "    # L2 regularization term on weights\n",
    "    # Increasing this value will make model more conservative.\n",
    "    # [default=1]\n",
    "    'lambda': 1,\n",
    "    \n",
    "    # Other parameters\n",
    "    'objective':'reg:squarederror',\n",
    "    'eval_metric': 'rmse'\n",
    "}\n",
    "\n",
    "model = xgb.train(params, dtrain, num_boost_round=999, evals=[(dval, \"Validation\")], early_stopping_rounds=30)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 153,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(110.840141, 15, 16)"
      ]
     },
     "execution_count": 153,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.best_score, model.best_iteration, model.best_ntree_limit"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "It works as expected, the training stopped after 10 rounds without improvement of the given metric."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.3. Tuning `max_depth` and `min_child_weight`\n",
    "* ***max_depth***: maximum depth of a tree. Increasing this value will make the model more complex and more likely to overfit\n",
    "* ***min_child_weight***: Defines the minimum sum of weights of all observations required in a child. Used to control over-fitting. Higher values prevent a model from learning relations which might be highly specific to the particular sample selected for a tree."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [],
   "source": [
    "gridsearch_params = [\n",
    "    (max_depth, min_child_weight)\n",
    "    for max_depth in [5, 6, 7, 8] \n",
    "    for min_child_weight in [1, 2, 3, 4, 5, 6]\n",
    "]\n",
    "param_names = ['max_depth', 'min_child_weight']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CV run for parameters:\n",
      "\tmax_depth: 5\tmin_child_weight: 1\n",
      "\tRMSE value: 130.68 for 26 rounds\n",
      "CV run for parameters:\n",
      "\tmax_depth: 5\tmin_child_weight: 2\n",
      "\tRMSE value: 129.30 for 40 rounds\n",
      "CV run for parameters:\n",
      "\tmax_depth: 5\tmin_child_weight: 3\n",
      "\tRMSE value: 131.03 for 43 rounds\n",
      "CV run for parameters:\n",
      "\tmax_depth: 5\tmin_child_weight: 4\n",
      "\tRMSE value: 132.12 for 37 rounds\n",
      "CV run for parameters:\n",
      "\tmax_depth: 5\tmin_child_weight: 5\n",
      "\tRMSE value: 131.71 for 51 rounds\n",
      "CV run for parameters:\n",
      "\tmax_depth: 5\tmin_child_weight: 6\n",
      "\tRMSE value: 132.07 for 37 rounds\n",
      "CV run for parameters:\n",
      "\tmax_depth: 6\tmin_child_weight: 1\n",
      "\tRMSE value: 131.72 for 16 rounds\n",
      "CV run for parameters:\n",
      "\tmax_depth: 6\tmin_child_weight: 2\n",
      "\tRMSE value: 130.59 for 21 rounds\n",
      "CV run for parameters:\n",
      "\tmax_depth: 6\tmin_child_weight: 3\n",
      "\tRMSE value: 132.31 for 31 rounds\n",
      "CV run for parameters:\n",
      "\tmax_depth: 6\tmin_child_weight: 4\n",
      "\tRMSE value: 131.89 for 31 rounds\n",
      "CV run for parameters:\n",
      "\tmax_depth: 6\tmin_child_weight: 5\n",
      "\tRMSE value: 131.62 for 48 rounds\n",
      "CV run for parameters:\n",
      "\tmax_depth: 6\tmin_child_weight: 6\n",
      "\tRMSE value: 131.77 for 20 rounds\n",
      "CV run for parameters:\n",
      "\tmax_depth: 7\tmin_child_weight: 1\n",
      "\tRMSE value: 131.52 for 20 rounds\n",
      "CV run for parameters:\n",
      "\tmax_depth: 7\tmin_child_weight: 2\n",
      "\tRMSE value: 130.40 for 17 rounds\n",
      "CV run for parameters:\n",
      "\tmax_depth: 7\tmin_child_weight: 3\n",
      "\tRMSE value: 132.88 for 15 rounds\n",
      "CV run for parameters:\n",
      "\tmax_depth: 7\tmin_child_weight: 4\n",
      "\tRMSE value: 131.85 for 29 rounds\n",
      "CV run for parameters:\n",
      "\tmax_depth: 7\tmin_child_weight: 5\n",
      "\tRMSE value: 131.85 for 27 rounds\n",
      "CV run for parameters:\n",
      "\tmax_depth: 7\tmin_child_weight: 6\n",
      "\tRMSE value: 131.58 for 32 rounds\n",
      "CV run for parameters:\n",
      "\tmax_depth: 8\tmin_child_weight: 1\n",
      "\tRMSE value: 130.27 for 35 rounds\n",
      "CV run for parameters:\n",
      "\tmax_depth: 8\tmin_child_weight: 2\n",
      "\tRMSE value: 131.77 for 40 rounds\n",
      "CV run for parameters:\n",
      "\tmax_depth: 8\tmin_child_weight: 3\n",
      "\tRMSE value: 132.66 for 19 rounds\n",
      "CV run for parameters:\n",
      "\tmax_depth: 8\tmin_child_weight: 4\n",
      "\tRMSE value: 133.03 for 23 rounds\n",
      "CV run for parameters:\n",
      "\tmax_depth: 8\tmin_child_weight: 5\n",
      "\tRMSE value: 132.70 for 16 rounds\n",
      "CV run for parameters:\n",
      "\tmax_depth: 8\tmin_child_weight: 6\n",
      "\tRMSE value: 131.15 for 19 rounds\n",
      "Best params['max_depth', 'min_child_weight']: [5, 2], RMSE value: 129.30\n"
     ]
    }
   ],
   "source": [
    "modeling.find_best_parameters(dtrain, params, gridsearch_params, param_names, 10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.4. Tuning `gamma`\n",
    "* ***gamma***: node is split only when the resulting split gives a positive reduction in the loss function. Gamma specifies the minimum loss reduction required to make a further partition on a leaf node of the tree. The larger gamma is, the more conservative the algorithm will be."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 154,
   "metadata": {},
   "outputs": [],
   "source": [
    "gridsearch_params = [0, 0.5, 1, 5, 10, 20, 30]\n",
    "param_names = ['gamma']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 155,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CV run for parameters:\n",
      "\tgamma: 0\n",
      "\tRMSE value: 129.30 for 40 rounds\n",
      "CV run for parameters:\n",
      "\tgamma: 0.5\n",
      "\tRMSE value: 129.30 for 40 rounds\n",
      "CV run for parameters:\n",
      "\tgamma: 1\n",
      "\tRMSE value: 129.30 for 40 rounds\n",
      "CV run for parameters:\n",
      "\tgamma: 5\n",
      "\tRMSE value: 129.30 for 40 rounds\n",
      "CV run for parameters:\n",
      "\tgamma: 10\n",
      "\tRMSE value: 129.30 for 40 rounds\n",
      "CV run for parameters:\n",
      "\tgamma: 20\n",
      "\tRMSE value: 129.30 for 40 rounds\n",
      "CV run for parameters:\n",
      "\tgamma: 30\n",
      "\tRMSE value: 129.30 for 40 rounds\n",
      "Best params['gamma']: [30], RMSE value: 129.30\n"
     ]
    }
   ],
   "source": [
    "modeling.find_best_parameters(dtrain, params, gridsearch_params, param_names, 20)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Changing `gamma` value does not seem to have an impact, this is weird. Even if the cross validation gives 30 as the best result, I will keep the default value as results are the same.\n",
    "\n",
    "### 3.5. Tuning `subsample` and `colsample_bytree`\n",
    "* ***subsample***: denotes the fraction of observations to be randomly samples for each tree. Lower values make the algorithm more conservative and prevents overfitting but too small values might lead to under-fitting.\n",
    "* ***colsample_bytree***: Subsample ratio of columns when constructing each tree"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {},
   "outputs": [],
   "source": [
    "gridsearch_params = [\n",
    "    (subsample, colsample_bytree)\n",
    "    for subsample in [i/10.0 for i in range(6, 11)]\n",
    "    for colsample_bytree in [i/10.0 for i in range(6, 11)]\n",
    "]\n",
    "\n",
    "param_names = ['subsample', 'colsample_bytree']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CV run for parameters:\n",
      "\tsubsample: 0.6\tcolsample_bytree: 0.6\n",
      "\tRMSE value: 134.76 for 17 rounds\n",
      "CV run for parameters:\n",
      "\tsubsample: 0.6\tcolsample_bytree: 0.7\n",
      "\tRMSE value: 135.19 for 24 rounds\n",
      "CV run for parameters:\n",
      "\tsubsample: 0.6\tcolsample_bytree: 0.8\n",
      "\tRMSE value: 135.18 for 20 rounds\n",
      "CV run for parameters:\n",
      "\tsubsample: 0.6\tcolsample_bytree: 0.9\n",
      "\tRMSE value: 135.81 for 35 rounds\n",
      "CV run for parameters:\n",
      "\tsubsample: 0.6\tcolsample_bytree: 1.0\n",
      "\tRMSE value: 134.29 for 40 rounds\n",
      "CV run for parameters:\n",
      "\tsubsample: 0.7\tcolsample_bytree: 0.6\n",
      "\tRMSE value: 133.40 for 21 rounds\n",
      "CV run for parameters:\n",
      "\tsubsample: 0.7\tcolsample_bytree: 0.7\n",
      "\tRMSE value: 135.77 for 16 rounds\n",
      "CV run for parameters:\n",
      "\tsubsample: 0.7\tcolsample_bytree: 0.8\n",
      "\tRMSE value: 132.75 for 28 rounds\n",
      "CV run for parameters:\n",
      "\tsubsample: 0.7\tcolsample_bytree: 0.9\n",
      "\tRMSE value: 133.24 for 34 rounds\n",
      "CV run for parameters:\n",
      "\tsubsample: 0.7\tcolsample_bytree: 1.0\n",
      "\tRMSE value: 132.87 for 41 rounds\n",
      "CV run for parameters:\n",
      "\tsubsample: 0.8\tcolsample_bytree: 0.6\n",
      "\tRMSE value: 133.14 for 46 rounds\n",
      "CV run for parameters:\n",
      "\tsubsample: 0.8\tcolsample_bytree: 0.7\n",
      "\tRMSE value: 132.87 for 49 rounds\n",
      "CV run for parameters:\n",
      "\tsubsample: 0.8\tcolsample_bytree: 0.8\n",
      "\tRMSE value: 133.52 for 36 rounds\n",
      "CV run for parameters:\n",
      "\tsubsample: 0.8\tcolsample_bytree: 0.9\n",
      "\tRMSE value: 133.27 for 52 rounds\n",
      "CV run for parameters:\n",
      "\tsubsample: 0.8\tcolsample_bytree: 1.0\n",
      "\tRMSE value: 133.00 for 31 rounds\n",
      "CV run for parameters:\n",
      "\tsubsample: 0.9\tcolsample_bytree: 0.6\n",
      "\tRMSE value: 132.64 for 40 rounds\n",
      "CV run for parameters:\n",
      "\tsubsample: 0.9\tcolsample_bytree: 0.7\n",
      "\tRMSE value: 132.93 for 48 rounds\n",
      "CV run for parameters:\n",
      "\tsubsample: 0.9\tcolsample_bytree: 0.8\n",
      "\tRMSE value: 131.85 for 50 rounds\n",
      "CV run for parameters:\n",
      "\tsubsample: 0.9\tcolsample_bytree: 0.9\n",
      "\tRMSE value: 131.82 for 40 rounds\n",
      "CV run for parameters:\n",
      "\tsubsample: 0.9\tcolsample_bytree: 1.0\n",
      "\tRMSE value: 133.66 for 52 rounds\n",
      "CV run for parameters:\n",
      "\tsubsample: 1.0\tcolsample_bytree: 0.6\n",
      "\tRMSE value: 132.06 for 41 rounds\n",
      "CV run for parameters:\n",
      "\tsubsample: 1.0\tcolsample_bytree: 0.7\n",
      "\tRMSE value: 130.17 for 58 rounds\n",
      "CV run for parameters:\n",
      "\tsubsample: 1.0\tcolsample_bytree: 0.8\n",
      "\tRMSE value: 129.39 for 44 rounds\n",
      "CV run for parameters:\n",
      "\tsubsample: 1.0\tcolsample_bytree: 0.9\n",
      "\tRMSE value: 131.33 for 39 rounds\n",
      "CV run for parameters:\n",
      "\tsubsample: 1.0\tcolsample_bytree: 1.0\n",
      "\tRMSE value: 129.30 for 40 rounds\n",
      "Best params['subsample', 'colsample_bytree']: [1.0, 1.0], RMSE value: 129.30\n"
     ]
    }
   ],
   "source": [
    "modeling.find_best_parameters(dtrain, params, gridsearch_params, param_names, 30)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.6. Tuning `alpha` and `lambda`\n",
    "* ***alpha***: denotes the fraction of observations to be randomly samples for each tree. Lower values make the algorithm more conservative and prevents overfitting but too small values might lead to under-fitting.\n",
    "* ***lambda***: Subsample ratio of columns when constructing each tree"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {},
   "outputs": [],
   "source": [
    "gridsearch_params = [\n",
    "    (alpha, reg_lambda)\n",
    "    for alpha in [0.02, 0.03, 0.04, 0.1, 1]\n",
    "    for reg_lambda in [1, 3, 5]\n",
    "]\n",
    "\n",
    "param_names = ['alpha', 'lambda']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CV run for parameters:\n",
      "\talpha: 0.02\tlambda: 1\n",
      "\tRMSE value: 129.30 for 40 rounds\n",
      "CV run for parameters:\n",
      "\talpha: 0.02\tlambda: 3\n",
      "\tRMSE value: 130.35 for 61 rounds\n",
      "CV run for parameters:\n",
      "\talpha: 0.02\tlambda: 5\n",
      "\tRMSE value: 129.41 for 71 rounds\n",
      "CV run for parameters:\n",
      "\talpha: 0.03\tlambda: 1\n",
      "\tRMSE value: 129.30 for 40 rounds\n",
      "CV run for parameters:\n",
      "\talpha: 0.03\tlambda: 3\n",
      "\tRMSE value: 130.35 for 61 rounds\n",
      "CV run for parameters:\n",
      "\talpha: 0.03\tlambda: 5\n",
      "\tRMSE value: 129.41 for 71 rounds\n",
      "CV run for parameters:\n",
      "\talpha: 0.04\tlambda: 1\n",
      "\tRMSE value: 129.30 for 40 rounds\n",
      "CV run for parameters:\n",
      "\talpha: 0.04\tlambda: 3\n",
      "\tRMSE value: 130.35 for 61 rounds\n",
      "CV run for parameters:\n",
      "\talpha: 0.04\tlambda: 5\n",
      "\tRMSE value: 129.41 for 71 rounds\n",
      "CV run for parameters:\n",
      "\talpha: 0.1\tlambda: 1\n",
      "\tRMSE value: 129.30 for 40 rounds\n",
      "CV run for parameters:\n",
      "\talpha: 0.1\tlambda: 3\n",
      "\tRMSE value: 130.34 for 61 rounds\n",
      "CV run for parameters:\n",
      "\talpha: 0.1\tlambda: 5\n",
      "\tRMSE value: 129.46 for 70 rounds\n",
      "CV run for parameters:\n",
      "\talpha: 1\tlambda: 1\n",
      "\tRMSE value: 129.69 for 40 rounds\n",
      "CV run for parameters:\n",
      "\talpha: 1\tlambda: 3\n",
      "\tRMSE value: 129.94 for 61 rounds\n",
      "CV run for parameters:\n",
      "\talpha: 1\tlambda: 5\n",
      "\tRMSE value: 130.00 for 37 rounds\n",
      "Best params['alpha', 'lambda']: [0.1, 1], RMSE value: 129.30\n"
     ]
    }
   ],
   "source": [
    "modeling.find_best_parameters(dtrain, params, gridsearch_params, param_names, 10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "It seems that `alpha` has no impact whereas `lambda` has a little one because there is a pattern:\n",
    "* RMSE = 129.30 when lambda=1\n",
    "* RMSE = 130.34 when lambda=3\n",
    "* RMSE = 129.41 when lambda=5\n",
    "\n",
    "### 3.7. Tuning `eta`\n",
    "* ***eta***: it corresponds to the learning rate. Lower values will increase our number of trees."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 148,
   "metadata": {},
   "outputs": [],
   "source": [
    "gridsearch_params = [.3, .2, .1, .05, .01, .005]\n",
    "param_names = ['eta']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CV run for parameters:\n",
      "\teta: 0.3\n",
      "\tRMSE value: 129.30 for 40 rounds\n",
      "CV run for parameters:\n",
      "\teta: 0.2\n",
      "\tRMSE value: 129.61 for 90 rounds\n",
      "CV run for parameters:\n",
      "\teta: 0.1\n",
      "\tRMSE value: 128.06 for 201 rounds\n",
      "CV run for parameters:\n",
      "\teta: 0.05\n",
      "\tRMSE value: 127.32 for 327 rounds\n",
      "CV run for parameters:\n",
      "\teta: 0.01\n",
      "\tRMSE value: 128.40 for 998 rounds\n",
      "CV run for parameters:\n",
      "\teta: 0.005\n",
      "\tRMSE value: 129.94 for 998 rounds\n",
      "Best params['eta']: [0.05], RMSE value: 127.32\n"
     ]
    }
   ],
   "source": [
    "modeling.find_best_parameters(dtrain, params, gridsearch_params, param_names, 10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.8. Putting it all together\n",
    "After all cross validation runs, our best parameters to use are:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 157,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'min_child_weight': 2,\n",
       " 'max_depth': 5,\n",
       " 'gamma': 0,\n",
       " 'eta': 0.05,\n",
       " 'subsample': 1.0,\n",
       " 'colsample_bytree': 1.0,\n",
       " 'alpha': 0.1,\n",
       " 'lambda': 1,\n",
       " 'objective': 'reg:squarederror',\n",
       " 'eval_metric': 'rmse'}"
      ]
     },
     "execution_count": 157,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "params"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's now fit and predict with such parameters. We will use 328 rounds (given by last cv run for `eta`)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 161,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0]\tValidation-rmse:189.412\n",
      "Will train until Validation-rmse hasn't improved in 10 rounds.\n",
      "[1]\tValidation-rmse:183.97\n",
      "[2]\tValidation-rmse:179.007\n",
      "[3]\tValidation-rmse:174.205\n",
      "[4]\tValidation-rmse:169.865\n",
      "[5]\tValidation-rmse:165.522\n",
      "[6]\tValidation-rmse:161.736\n",
      "[7]\tValidation-rmse:158.058\n",
      "[8]\tValidation-rmse:154.753\n",
      "[9]\tValidation-rmse:151.597\n",
      "[10]\tValidation-rmse:148.662\n",
      "[11]\tValidation-rmse:145.911\n",
      "[12]\tValidation-rmse:143.275\n",
      "[13]\tValidation-rmse:140.834\n",
      "[14]\tValidation-rmse:138.393\n",
      "[15]\tValidation-rmse:136.453\n",
      "[16]\tValidation-rmse:134.603\n",
      "[17]\tValidation-rmse:132.764\n",
      "[18]\tValidation-rmse:131.143\n",
      "[19]\tValidation-rmse:129.659\n",
      "[20]\tValidation-rmse:128.128\n",
      "[21]\tValidation-rmse:126.657\n",
      "[22]\tValidation-rmse:125.417\n",
      "[23]\tValidation-rmse:124.262\n",
      "[24]\tValidation-rmse:123.198\n",
      "[25]\tValidation-rmse:121.988\n",
      "[26]\tValidation-rmse:121.007\n",
      "[27]\tValidation-rmse:120.102\n",
      "[28]\tValidation-rmse:119.151\n",
      "[29]\tValidation-rmse:118.391\n",
      "[30]\tValidation-rmse:117.549\n",
      "[31]\tValidation-rmse:116.817\n",
      "[32]\tValidation-rmse:116.163\n",
      "[33]\tValidation-rmse:115.608\n",
      "[34]\tValidation-rmse:115.132\n",
      "[35]\tValidation-rmse:114.678\n",
      "[36]\tValidation-rmse:114.193\n",
      "[37]\tValidation-rmse:113.768\n",
      "[38]\tValidation-rmse:113.356\n",
      "[39]\tValidation-rmse:112.941\n",
      "[40]\tValidation-rmse:112.494\n",
      "[41]\tValidation-rmse:112.123\n",
      "[42]\tValidation-rmse:111.696\n",
      "[43]\tValidation-rmse:111.331\n",
      "[44]\tValidation-rmse:110.958\n",
      "[45]\tValidation-rmse:110.704\n",
      "[46]\tValidation-rmse:110.186\n",
      "[47]\tValidation-rmse:109.897\n",
      "[48]\tValidation-rmse:109.649\n",
      "[49]\tValidation-rmse:109.424\n",
      "[50]\tValidation-rmse:109.328\n",
      "[51]\tValidation-rmse:109.122\n",
      "[52]\tValidation-rmse:109.048\n",
      "[53]\tValidation-rmse:108.823\n",
      "[54]\tValidation-rmse:108.823\n",
      "[55]\tValidation-rmse:108.839\n",
      "[56]\tValidation-rmse:108.701\n",
      "[57]\tValidation-rmse:108.773\n",
      "[58]\tValidation-rmse:108.862\n",
      "[59]\tValidation-rmse:109.053\n",
      "[60]\tValidation-rmse:108.907\n",
      "[61]\tValidation-rmse:108.671\n",
      "[62]\tValidation-rmse:108.795\n",
      "[63]\tValidation-rmse:108.611\n",
      "[64]\tValidation-rmse:108.791\n",
      "[65]\tValidation-rmse:108.584\n",
      "[66]\tValidation-rmse:108.767\n",
      "[67]\tValidation-rmse:108.424\n",
      "[68]\tValidation-rmse:108.297\n",
      "[69]\tValidation-rmse:108.25\n",
      "[70]\tValidation-rmse:108.271\n",
      "[71]\tValidation-rmse:108.154\n",
      "[72]\tValidation-rmse:107.996\n",
      "[73]\tValidation-rmse:107.932\n",
      "[74]\tValidation-rmse:107.683\n",
      "[75]\tValidation-rmse:107.87\n",
      "[76]\tValidation-rmse:107.797\n",
      "[77]\tValidation-rmse:107.747\n",
      "[78]\tValidation-rmse:107.67\n",
      "[79]\tValidation-rmse:107.61\n",
      "[80]\tValidation-rmse:107.855\n",
      "[81]\tValidation-rmse:107.744\n",
      "[82]\tValidation-rmse:107.696\n",
      "[83]\tValidation-rmse:107.597\n",
      "[84]\tValidation-rmse:107.552\n",
      "[85]\tValidation-rmse:107.464\n",
      "[86]\tValidation-rmse:107.329\n",
      "[87]\tValidation-rmse:107.289\n",
      "[88]\tValidation-rmse:107.617\n",
      "[89]\tValidation-rmse:107.578\n",
      "[90]\tValidation-rmse:107.968\n",
      "[91]\tValidation-rmse:108.442\n",
      "[92]\tValidation-rmse:108.366\n",
      "[93]\tValidation-rmse:108.284\n",
      "[94]\tValidation-rmse:108.627\n",
      "[95]\tValidation-rmse:108.399\n",
      "[96]\tValidation-rmse:108.346\n",
      "[97]\tValidation-rmse:108.245\n",
      "Stopping. Best iteration:\n",
      "[87]\tValidation-rmse:107.289\n",
      "\n"
     ]
    }
   ],
   "source": [
    "best_model = xgb.train(params, dtrain, num_boost_round=328, evals=[(dval, \"Validation\")], early_stopping_rounds=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 168,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "RMSE for XGBoost tuned with internal cv on train: 114.85\n"
     ]
    }
   ],
   "source": [
    "y_xgb_run_py_api_train = best_model.predict(dtrain)\n",
    "print(\"RMSE for XGBoost tuned with internal cv on train: {:.2f}\".format(np.sqrt(mean_squared_error(y_train, y_xgb_run_py_api_train))))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 162,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "RMSE for XGBoost tuned with internal cv on test: 104.42\n"
     ]
    }
   ],
   "source": [
    "y_xgb_run_py_api = best_model.predict(dtest)\n",
    "print(\"RMSE for XGBoost tuned with internal cv on test: {:.2f}\".format(np.sqrt(mean_squared_error(y_test, y_xgb_run_py_api))))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n",
    "---\n",
    "# 4. Another RandomizedSearch with more values\n",
    "## 4.1. Build parameter grid and launch search\n",
    "Let's try to increase the number of estimators and explore more values.\n",
    "\n",
    "As the `validation` dataset is not used with RandomSearch, we are losing some interesting data. The idea here is to rebuild a bigger `train` dataset and use it for the RandomSearch."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X_train2.shape is (57863, 153)\n",
      "y_train2.shape is (57863, 1)\n"
     ]
    }
   ],
   "source": [
    "y_val.columns = ['train_price']\n",
    "X_train2 = pd.concat([X_train, X_val], axis=0)\n",
    "y_train2 = pd.concat([y_train, y_val], axis=0)\n",
    "print(\"X_train2.shape is {}\".format(X_train2.shape))\n",
    "print(\"y_train2.shape is {}\".format(y_train2.shape))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 100 candidates, totalling 500 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed: 20.4min\n",
      "[Parallel(n_jobs=-1)]: Done 184 tasks      | elapsed: 105.0min\n",
      "[Parallel(n_jobs=-1)]: Done 434 tasks      | elapsed: 279.4min\n",
      "[Parallel(n_jobs=-1)]: Done 500 out of 500 | elapsed: 317.1min finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "RMSE for tuned and optimized xgboost model on train: 30.43\n",
      "RMSE for tuned and optimized xgboost model on test: 101.88\n"
     ]
    }
   ],
   "source": [
    "# Parameters of xgboost that may change\n",
    "param_grid = {\n",
    "    'n_estimators': [100, 300, 500, 700, 800], # Number of estimators (weak learners)\n",
    "    \n",
    "    # Defines the minimum sum of weights of all observations required in a child.\n",
    "    # Used to control over-fitting. Higher values prevent a model from learning relations which might be highly specific\n",
    "    # to the particular sample selected for a tree.\n",
    "    # [default=1]\n",
    "    'min_child_weight': [1, 2, 3, 4, 5, 6],\n",
    "    \n",
    "    # Maximum depth of a tree. Increasing this value will make the model more complex and more likely to overfit\n",
    "    'max_depth': [5, 6, 7, 8],\n",
    "    \n",
    "    # Gamma: A node is split only when the resulting split gives a positive reduction in the loss function.\n",
    "    # Gamma specifies the minimum loss reduction required to make a further partition on a leaf node of the tree.\n",
    "    # Makes the algorithm conservative.\n",
    "    'gamma': [i/10.0 for i in range(0, 7)],\n",
    "    \n",
    "    # Learning rate\n",
    "    'learning_rate': [.3, .2, .1, .05, .01, .005],\n",
    "    \n",
    "    # Denotes the fraction of observations to be randomly samples for each tree.\n",
    "    # Lower values make the algorithm more conservative and prevents overfitting but too small values might lead to \n",
    "    # under-fitting. Typical values: 0.5-1\n",
    "    # [default=1]\n",
    "    'subsample': [i/10.0 for i in range(6, 11)],\n",
    "    'colsample_bytree': [i/10.0 for i in range(6, 11)], # subsample ratio of columns when constructing each tree (range is [0, 1])\n",
    "    \n",
    "    # L1 regularization term on weights. Increasing this value will make model more conservative\n",
    "    # Can be used in case of very high dimensionality so that the algorithm runs faster when implemented\n",
    "    # [default=0]\n",
    "    'reg_alpha': [0.02, 0.03, 0.04, 0.1, 1], \n",
    "    \n",
    "    # L2 regularization term on weights\n",
    "    # Increasing this value will make model more conservative.\n",
    "    # [default=1]\n",
    "    'reg_lambda': [1, 3, 5]\n",
    "}\n",
    "\n",
    "xgb_randomsearch2 = modeling.build_xgb_random_search(param_grid, num_iters=100)\n",
    "y_xgbrs_run2 = modeling.fit_and_run_pipeline(xgb_randomsearch2, 'tuned and optimized xgboost', X_train, y_train, X_test, y_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "***Observation:*** there is still overfitting but it seems we have found a better combination for parameters.  \n",
    "Let's display what were our best estimator."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "XGBRegressor(base_score=0.5, booster='gbtree', colsample_bylevel=1,\n",
       "             colsample_bynode=1, colsample_bytree=0.8, eval_metric='rmse',\n",
       "             gamma=0.6, importance_type='gain', learning_rate=0.1,\n",
       "             max_delta_step=0, max_depth=7, min_child_weight=2, missing=None,\n",
       "             n_estimators=800, n_jobs=1, nthread=None,\n",
       "             objective='reg:squarederror', random_state=0, reg_alpha=0.1,\n",
       "             reg_lambda=3, scale_pos_weight=1, seed=42, silent=None,\n",
       "             subsample=1.0, verbose=False, verbosity=1)"
      ]
     },
     "execution_count": 108,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "xgb_randomsearch2.best_estimator_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4.2. Refit with earlystopping\n",
    "First, we do not activate the earlystopping."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = xgb_randomsearch2.best_estimator_.fit(X_train, y_train, verbose=False,\n",
    "                                              eval_set=[(X_train, y_train), (X_val, y_val)], eval_metric='rmse')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Display learning curves"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 137,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "viz.plot_learning_curves(model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 138,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_es = xgb_randomsearch2.best_estimator_.fit(X_train, y_train, verbose=False,\n",
    "                                              eval_set=[(X_train, y_train), (X_val, y_val)], eval_metric='rmse',\n",
    "                                             early_stopping_rounds=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 139,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "viz.plot_learning_curves(model_es)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 169,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "RMSE for XGBoost with early stopping model on train: 103.43\n"
     ]
    }
   ],
   "source": [
    "y_xgbrs_run_with_es_train = model_es.predict(X_train)\n",
    "print(\"RMSE for XGBoost with early stopping model on train: {:.2f}\".format(np.sqrt(mean_squared_error(y_train, y_xgbrs_run_with_es_train))))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 142,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "RMSE for XGBoost with early stopping model on test: 101.09\n"
     ]
    }
   ],
   "source": [
    "y_xgbrs_run_with_es = model_es.predict(X_test)\n",
    "print(\"RMSE for XGBoost with early stopping model on test: {:.2f}\".format(np.sqrt(mean_squared_error(y_test, y_xgbrs_run_with_es))))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The early stopping function works as expected and we can clearly see that there is no need to have a lot of estimators (trees), the more we add the more it overfits without improvement on validation set. Results are even slightly better."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "## 5. Results comparison\n",
    "Let's concatenate all our predictions into a new dataframe so that we can compare them. We have:\n",
    "* ***y_true***: the real price\n",
    "* ***y_xgbrs_run_light***: (xgboost sklearn wrapper) prediction with RandomizedSearch 'light' (400 fits)\n",
    "* ***y_xgbrs_run_full***: (xgboost sklearn wrapper) prediction with RandomizedSearch and more values to explore (500 fits)\n",
    "* ***y_xgbrs_run_with_es***: (xgboost sklearn wrapper) the best model found with run full + earlystopping\n",
    "* ***y_xgb_run_py_api***: (xgboost internal python api) best model found after parameters tuning then run with earlystopping"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 165,
   "metadata": {},
   "outputs": [
    {
     "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>y_true</th>\n",
       "      <th>y_xgbrs_run_light</th>\n",
       "      <th>y_xgbrs_run_full</th>\n",
       "      <th>y_xgbrs_run_with_es</th>\n",
       "      <th>y_xgb_run_py_api</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>59.0</td>\n",
       "      <td>63.762135</td>\n",
       "      <td>62.879246</td>\n",
       "      <td>65.232857</td>\n",
       "      <td>67.108482</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>90.0</td>\n",
       "      <td>80.044769</td>\n",
       "      <td>85.548782</td>\n",
       "      <td>74.711510</td>\n",
       "      <td>77.864044</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>46.0</td>\n",
       "      <td>55.033432</td>\n",
       "      <td>54.515701</td>\n",
       "      <td>59.607632</td>\n",
       "      <td>58.642143</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>67.0</td>\n",
       "      <td>90.333168</td>\n",
       "      <td>85.936150</td>\n",
       "      <td>95.750923</td>\n",
       "      <td>98.538551</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>59.0</td>\n",
       "      <td>84.442856</td>\n",
       "      <td>77.867317</td>\n",
       "      <td>81.731850</td>\n",
       "      <td>88.702660</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>95.0</td>\n",
       "      <td>108.445274</td>\n",
       "      <td>101.893532</td>\n",
       "      <td>104.948822</td>\n",
       "      <td>110.145081</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>90.0</td>\n",
       "      <td>78.571976</td>\n",
       "      <td>76.608368</td>\n",
       "      <td>77.621010</td>\n",
       "      <td>80.645348</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>119.0</td>\n",
       "      <td>514.855530</td>\n",
       "      <td>536.258850</td>\n",
       "      <td>524.725891</td>\n",
       "      <td>404.656189</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>90.0</td>\n",
       "      <td>130.405289</td>\n",
       "      <td>103.199371</td>\n",
       "      <td>121.328766</td>\n",
       "      <td>129.756149</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>65.0</td>\n",
       "      <td>77.177063</td>\n",
       "      <td>71.133011</td>\n",
       "      <td>68.866974</td>\n",
       "      <td>74.254250</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   y_true  y_xgbrs_run_light  y_xgbrs_run_full  y_xgbrs_run_with_es  \\\n",
       "0    59.0          63.762135         62.879246            65.232857   \n",
       "1    90.0          80.044769         85.548782            74.711510   \n",
       "2    46.0          55.033432         54.515701            59.607632   \n",
       "3    67.0          90.333168         85.936150            95.750923   \n",
       "4    59.0          84.442856         77.867317            81.731850   \n",
       "5    95.0         108.445274        101.893532           104.948822   \n",
       "6    90.0          78.571976         76.608368            77.621010   \n",
       "7   119.0         514.855530        536.258850           524.725891   \n",
       "8    90.0         130.405289        103.199371           121.328766   \n",
       "9    65.0          77.177063         71.133011            68.866974   \n",
       "\n",
       "   y_xgb_run_py_api  \n",
       "0         67.108482  \n",
       "1         77.864044  \n",
       "2         58.642143  \n",
       "3         98.538551  \n",
       "4         88.702660  \n",
       "5        110.145081  \n",
       "6         80.645348  \n",
       "7        404.656189  \n",
       "8        129.756149  \n",
       "9         74.254250  "
      ]
     },
     "execution_count": 165,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results_df = pd.concat([y_test, \n",
    "                        pd.DataFrame(data=y_xgbrs_run1), \n",
    "                        pd.DataFrame(data=y_xgbrs_run2), \n",
    "                        pd.DataFrame(data=y_xgbrs_run_with_es),\n",
    "                        pd.DataFrame(data=y_xgb_run_py_api)],\n",
    "                       axis=1)\n",
    "results_df.columns = ['y_true', 'y_xgbrs_run_light', 'y_xgbrs_run_full', 'y_xgbrs_run_with_es', 'y_xgb_run_py_api']\n",
    "results_df.head(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 166,
   "metadata": {},
   "outputs": [],
   "source": [
    "results_df = modeling.classify_results(results_df, 'y_xgbrs_run_light')\n",
    "results_df = modeling.classify_results(results_df, 'y_xgbrs_run_full')\n",
    "results_df = modeling.classify_results(results_df, 'y_xgbrs_run_with_es')\n",
    "results_df = modeling.classify_results(results_df, 'y_xgb_run_py_api')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 167,
   "metadata": {},
   "outputs": [
    {
     "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>y_true</th>\n",
       "      <th>y_xgbrs_run_light</th>\n",
       "      <th>y_xgbrs_run_full</th>\n",
       "      <th>y_xgbrs_run_with_es</th>\n",
       "      <th>y_xgb_run_py_api</th>\n",
       "      <th>y_xgbrs_run_light_perc_diff</th>\n",
       "      <th>y_xgbrs_run_light_perc_diff_class</th>\n",
       "      <th>y_xgbrs_run_full_perc_diff</th>\n",
       "      <th>y_xgbrs_run_full_perc_diff_class</th>\n",
       "      <th>y_xgbrs_run_with_es_perc_diff</th>\n",
       "      <th>y_xgbrs_run_with_es_perc_diff_class</th>\n",
       "      <th>y_xgb_run_py_api_perc_diff</th>\n",
       "      <th>y_xgb_run_py_api_perc_diff_class</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>59.0</td>\n",
       "      <td>63.762135</td>\n",
       "      <td>62.879246</td>\n",
       "      <td>65.232857</td>\n",
       "      <td>67.108482</td>\n",
       "      <td>8.07</td>\n",
       "      <td>3-good (5% &lt; x &lt; 10%)</td>\n",
       "      <td>6.57</td>\n",
       "      <td>3-good (5% &lt; x &lt; 10%)</td>\n",
       "      <td>10.56</td>\n",
       "      <td>4-acceptable (10% &lt; x &lt; 20%)</td>\n",
       "      <td>13.74</td>\n",
       "      <td>4-acceptable (10% &lt; x &lt; 20%)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>90.0</td>\n",
       "      <td>80.044769</td>\n",
       "      <td>85.548782</td>\n",
       "      <td>74.711510</td>\n",
       "      <td>77.864044</td>\n",
       "      <td>11.06</td>\n",
       "      <td>4-acceptable (10% &lt; x &lt; 20%)</td>\n",
       "      <td>4.95</td>\n",
       "      <td>2-pretty good (2% &lt; x &lt; 5%)</td>\n",
       "      <td>16.99</td>\n",
       "      <td>4-acceptable (10% &lt; x &lt; 20%)</td>\n",
       "      <td>13.48</td>\n",
       "      <td>4-acceptable (10% &lt; x &lt; 20%)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>46.0</td>\n",
       "      <td>55.033432</td>\n",
       "      <td>54.515701</td>\n",
       "      <td>59.607632</td>\n",
       "      <td>58.642143</td>\n",
       "      <td>19.64</td>\n",
       "      <td>4-acceptable (10% &lt; x &lt; 20%)</td>\n",
       "      <td>18.51</td>\n",
       "      <td>4-acceptable (10% &lt; x &lt; 20%)</td>\n",
       "      <td>29.58</td>\n",
       "      <td>5-bad (20% &lt; x &lt; 40%)</td>\n",
       "      <td>27.48</td>\n",
       "      <td>5-bad (20% &lt; x &lt; 40%)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>67.0</td>\n",
       "      <td>90.333168</td>\n",
       "      <td>85.936150</td>\n",
       "      <td>95.750923</td>\n",
       "      <td>98.538551</td>\n",
       "      <td>34.83</td>\n",
       "      <td>5-bad (20% &lt; x &lt; 40%)</td>\n",
       "      <td>28.26</td>\n",
       "      <td>5-bad (20% &lt; x &lt; 40%)</td>\n",
       "      <td>42.91</td>\n",
       "      <td>6-awful (&gt; 40%)</td>\n",
       "      <td>47.07</td>\n",
       "      <td>6-awful (&gt; 40%)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>59.0</td>\n",
       "      <td>84.442856</td>\n",
       "      <td>77.867317</td>\n",
       "      <td>81.731850</td>\n",
       "      <td>88.702660</td>\n",
       "      <td>43.12</td>\n",
       "      <td>6-awful (&gt; 40%)</td>\n",
       "      <td>31.98</td>\n",
       "      <td>5-bad (20% &lt; x &lt; 40%)</td>\n",
       "      <td>38.53</td>\n",
       "      <td>5-bad (20% &lt; x &lt; 40%)</td>\n",
       "      <td>50.34</td>\n",
       "      <td>6-awful (&gt; 40%)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>95.0</td>\n",
       "      <td>108.445274</td>\n",
       "      <td>101.893532</td>\n",
       "      <td>104.948822</td>\n",
       "      <td>110.145081</td>\n",
       "      <td>14.15</td>\n",
       "      <td>4-acceptable (10% &lt; x &lt; 20%)</td>\n",
       "      <td>7.26</td>\n",
       "      <td>3-good (5% &lt; x &lt; 10%)</td>\n",
       "      <td>10.47</td>\n",
       "      <td>4-acceptable (10% &lt; x &lt; 20%)</td>\n",
       "      <td>15.94</td>\n",
       "      <td>4-acceptable (10% &lt; x &lt; 20%)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>90.0</td>\n",
       "      <td>78.571976</td>\n",
       "      <td>76.608368</td>\n",
       "      <td>77.621010</td>\n",
       "      <td>80.645348</td>\n",
       "      <td>12.70</td>\n",
       "      <td>4-acceptable (10% &lt; x &lt; 20%)</td>\n",
       "      <td>14.88</td>\n",
       "      <td>4-acceptable (10% &lt; x &lt; 20%)</td>\n",
       "      <td>13.75</td>\n",
       "      <td>4-acceptable (10% &lt; x &lt; 20%)</td>\n",
       "      <td>10.39</td>\n",
       "      <td>4-acceptable (10% &lt; x &lt; 20%)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>119.0</td>\n",
       "      <td>514.855530</td>\n",
       "      <td>536.258850</td>\n",
       "      <td>524.725891</td>\n",
       "      <td>404.656189</td>\n",
       "      <td>332.65</td>\n",
       "      <td>6-awful (&gt; 40%)</td>\n",
       "      <td>350.64</td>\n",
       "      <td>6-awful (&gt; 40%)</td>\n",
       "      <td>340.95</td>\n",
       "      <td>6-awful (&gt; 40%)</td>\n",
       "      <td>240.05</td>\n",
       "      <td>6-awful (&gt; 40%)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>90.0</td>\n",
       "      <td>130.405289</td>\n",
       "      <td>103.199371</td>\n",
       "      <td>121.328766</td>\n",
       "      <td>129.756149</td>\n",
       "      <td>44.89</td>\n",
       "      <td>6-awful (&gt; 40%)</td>\n",
       "      <td>14.67</td>\n",
       "      <td>4-acceptable (10% &lt; x &lt; 20%)</td>\n",
       "      <td>34.81</td>\n",
       "      <td>5-bad (20% &lt; x &lt; 40%)</td>\n",
       "      <td>44.17</td>\n",
       "      <td>6-awful (&gt; 40%)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>65.0</td>\n",
       "      <td>77.177063</td>\n",
       "      <td>71.133011</td>\n",
       "      <td>68.866974</td>\n",
       "      <td>74.254250</td>\n",
       "      <td>18.73</td>\n",
       "      <td>4-acceptable (10% &lt; x &lt; 20%)</td>\n",
       "      <td>9.44</td>\n",
       "      <td>3-good (5% &lt; x &lt; 10%)</td>\n",
       "      <td>5.95</td>\n",
       "      <td>3-good (5% &lt; x &lt; 10%)</td>\n",
       "      <td>14.24</td>\n",
       "      <td>4-acceptable (10% &lt; x &lt; 20%)</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   y_true  y_xgbrs_run_light  y_xgbrs_run_full  y_xgbrs_run_with_es  \\\n",
       "0    59.0          63.762135         62.879246            65.232857   \n",
       "1    90.0          80.044769         85.548782            74.711510   \n",
       "2    46.0          55.033432         54.515701            59.607632   \n",
       "3    67.0          90.333168         85.936150            95.750923   \n",
       "4    59.0          84.442856         77.867317            81.731850   \n",
       "5    95.0         108.445274        101.893532           104.948822   \n",
       "6    90.0          78.571976         76.608368            77.621010   \n",
       "7   119.0         514.855530        536.258850           524.725891   \n",
       "8    90.0         130.405289        103.199371           121.328766   \n",
       "9    65.0          77.177063         71.133011            68.866974   \n",
       "\n",
       "   y_xgb_run_py_api  y_xgbrs_run_light_perc_diff  \\\n",
       "0         67.108482                         8.07   \n",
       "1         77.864044                        11.06   \n",
       "2         58.642143                        19.64   \n",
       "3         98.538551                        34.83   \n",
       "4         88.702660                        43.12   \n",
       "5        110.145081                        14.15   \n",
       "6         80.645348                        12.70   \n",
       "7        404.656189                       332.65   \n",
       "8        129.756149                        44.89   \n",
       "9         74.254250                        18.73   \n",
       "\n",
       "  y_xgbrs_run_light_perc_diff_class  y_xgbrs_run_full_perc_diff  \\\n",
       "0             3-good (5% < x < 10%)                        6.57   \n",
       "1      4-acceptable (10% < x < 20%)                        4.95   \n",
       "2      4-acceptable (10% < x < 20%)                       18.51   \n",
       "3             5-bad (20% < x < 40%)                       28.26   \n",
       "4                   6-awful (> 40%)                       31.98   \n",
       "5      4-acceptable (10% < x < 20%)                        7.26   \n",
       "6      4-acceptable (10% < x < 20%)                       14.88   \n",
       "7                   6-awful (> 40%)                      350.64   \n",
       "8                   6-awful (> 40%)                       14.67   \n",
       "9      4-acceptable (10% < x < 20%)                        9.44   \n",
       "\n",
       "  y_xgbrs_run_full_perc_diff_class  y_xgbrs_run_with_es_perc_diff  \\\n",
       "0            3-good (5% < x < 10%)                          10.56   \n",
       "1      2-pretty good (2% < x < 5%)                          16.99   \n",
       "2     4-acceptable (10% < x < 20%)                          29.58   \n",
       "3            5-bad (20% < x < 40%)                          42.91   \n",
       "4            5-bad (20% < x < 40%)                          38.53   \n",
       "5            3-good (5% < x < 10%)                          10.47   \n",
       "6     4-acceptable (10% < x < 20%)                          13.75   \n",
       "7                  6-awful (> 40%)                         340.95   \n",
       "8     4-acceptable (10% < x < 20%)                          34.81   \n",
       "9            3-good (5% < x < 10%)                           5.95   \n",
       "\n",
       "  y_xgbrs_run_with_es_perc_diff_class  y_xgb_run_py_api_perc_diff  \\\n",
       "0        4-acceptable (10% < x < 20%)                       13.74   \n",
       "1        4-acceptable (10% < x < 20%)                       13.48   \n",
       "2               5-bad (20% < x < 40%)                       27.48   \n",
       "3                     6-awful (> 40%)                       47.07   \n",
       "4               5-bad (20% < x < 40%)                       50.34   \n",
       "5        4-acceptable (10% < x < 20%)                       15.94   \n",
       "6        4-acceptable (10% < x < 20%)                       10.39   \n",
       "7                     6-awful (> 40%)                      240.05   \n",
       "8               5-bad (20% < x < 40%)                       44.17   \n",
       "9               3-good (5% < x < 10%)                       14.24   \n",
       "\n",
       "  y_xgb_run_py_api_perc_diff_class  \n",
       "0     4-acceptable (10% < x < 20%)  \n",
       "1     4-acceptable (10% < x < 20%)  \n",
       "2            5-bad (20% < x < 40%)  \n",
       "3                  6-awful (> 40%)  \n",
       "4                  6-awful (> 40%)  \n",
       "5     4-acceptable (10% < x < 20%)  \n",
       "6     4-acceptable (10% < x < 20%)  \n",
       "7                  6-awful (> 40%)  \n",
       "8                  6-awful (> 40%)  \n",
       "9     4-acceptable (10% < x < 20%)  "
      ]
     },
     "execution_count": 167,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results_df.head(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### RMSE and classification table results\n",
    "As a reminder and to have a better view on improvement and progression, previous results are given here as well.  \n",
    "\n",
    "| Model              | RMSE value on train | RMSE value on test | Nb class 1-very accurate (< 2%) | Nb class 2-pretty good (2% < x < 5%) | Nb class 3-good (5% < x < 10%) | Nb class 4-acceptable (10% < x < 20%) | Nb class 5-bad (20% < x < 40%) | Nb class 6-awful (> 40%) |\n",
    "|--------------------|---------------------|--------------------|--------------------|--------------------|--------------------|--------------------|--------------------|--------------------|\n",
    "| Naive              | N/A                 | 138.02             | 148             | 240             | 453             | 727             | 1280             | 3582             |\n",
    "| Linear Regression  | 141.72              | 113.43             | 219             | 326             | 526             | 1086             | 1696             | 2577             |\n",
    "| Basic DecisionTree | 104.85              | 186.95             | 273             | 417             | 637             | 1161             | 1788             | 2154             |\n",
    "| Tuned DecisionTree | 125.43              | 114.46             | 301             | 393             | 632             | 1120             | 1847             | 2137             |\n",
    "| XGBoost (1st try)  | 40.59               | 102.73             | 347             | 515             | 864             | 1409             | 1816             | 1479             |\n",
    "| XGBoost (RS Light) | 90.87               | 102.38             | 332             | 483             | 837             | 1402             | 1854             | 1522             |\n",
    "| XGBoost (RS Full)  | 30.43               | 101.88             | 331             | 518             | 824             | 1444             | 1821             | 1492             |\n",
    "| XGBoost (Full + ES)| 103.43              | 101.09             | 312             | 483             | 735             | 1320             | 1887             | 1693             |\n",
    "| XGBoost (Py API )  | 114.85              | 104.42             | 300             | 453             | 745             | 1287             | 1847             | 1798             |\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 174,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x720 with 4 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "model_names = ['xgbrs_run_light', 'xgbrs_run_full', 'xgbrs_run_with_es', 'xgb_run_py_api']\n",
    "viz.plot_classification_results_pie(model_names, results_df)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "\n",
    "# CONCLUSION\n",
    "In spite of all the efforts made for parameters tuning (and there are a lot with XGBoost), I was unable to improve significantly the results I got with the first try and even if I found a combination that slightly improved the RMSE value, when analyzing in more details the first XGBoost remains the best choice as it is with it that most of the prices are near the truth.  \n",
    "Even if it is quite frustrating not having improved the results, let's recall that the first goal of this project was to show our ability to follow the Datascience process CRISP-DM and this is what has been done:\n",
    "1. `Data Understanding` in the [1st notebook](1_Data_Understanding.ipynb)\n",
    "2. `Business Understanding` in the [2nd one](2_Business_Understanding.ipynb) with also some checks against the online service.\n",
    "3. `Data preparation` [here](3_Data_Preparation.ipynb)\n",
    "4. `Modeling`, tuning and `Evaluation` (this notebook and the [previous one](4_Modeling.ipynb))\n",
    "\n",
    "It remains the `Deployment` (or `Exposition`) phase that is done through this [blog post](https://nidragedd.github.io/things-you-should-now-before-visiting-Paris/).  \n",
    "And let's say that to see the glass half full, we were able to build a model with 80% price prediction accuracy half the time. Naive model was 80% precision 25% of the time and linear regression 33% of the time.\n",
    "\n",
    "To go further and still try to improve, here are some things one could do:\n",
    "* Try Bayesian Optimization with [scikit-optimize](https://scikit-optimize.github.io/notebooks/sklearn-gridsearchcv-replacement.html) to provide search space and see if it gives better results.\n",
    "* Try [Catboost](https://catboost.ai/) from the russian company Yandex (as we have a lot of categorical features) and they say that there is less parameter tuning.\n",
    "* Check if dimensionnality reduction (with PCA) could help us.\n",
    "* Perform feature engineering by building new features.\n",
    "* Create ensemble of models, stack them to average the results."
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [conda env:udadsnd-p4] *",
   "language": "python",
   "name": "conda-env-udadsnd-p4-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}