docs/source/examples/fklearn_overview.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# FKLearn Tutorial: \n",
    "\n",
    "* <font size=\"4\"> FKlearn is nubank's functional library for Machine Learning (https://github.com/nubank/fklearn) </font>\n",
    "\n",
    "<img src=\"imgs/fklearn-github.png\">\n",
    "\n",
    "* <font size=\"4\"> It was created with the idea of **scaling machine learning** through the company by standardizing model development and implementing an easy interface to allow all users to develop the best practices on Machine Learning </font>\n",
    "\n",
    "\n",
    "* <font size=\"4\"> Currently powering more than **30 models in production** </font>\n",
    "\n",
    "\n",
    "* <font size=\"4\"> FKLearn was created having **4 principles** that guided it's development: </font>\n",
    "\n",
    "<img src=\"imgs/fklearn-principles.png\">\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# Input Analysis"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "## Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:16:43.982589Z",
     "start_time": "2019-09-26T20:16:43.019533Z"
    },
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "from matplotlib import pyplot as plt\n",
    "import seaborn as sns\n",
    "import matplotlib\n",
    "sns.set_style(\"whitegrid\")\n",
    "sns.set_palette(\"husl\")\n",
    "\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "##  Input Dataset\n",
    "\n",
    "* <font size=\"4\"> This dataset was created with simulated data about users **spend behavior on Credit Card** </font>\n",
    "\n",
    "* <font size=\"4\"> The model target is the average spend of the next 2 months and we created several features that are related to the target </font>\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:18:09.030706Z",
     "start_time": "2019-09-26T20:18:08.720385Z"
    },
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [],
   "source": [
    "# Generate this dataset using the FKLearn Tutorial Dataset.ipynb notebook\n",
    "df = pd.read_csv(\"fklearn-tutorial-input-dataset.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:18:10.725214Z",
     "start_time": "2019-09-26T20:18:10.692078Z"
    },
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [],
   "source": [
    "df['month_date'] = pd.to_datetime(df.month_date)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:18:11.795039Z",
     "start_time": "2019-09-26T20:18:11.774834Z"
    },
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "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>id</th>\n",
       "      <th>month</th>\n",
       "      <th>income</th>\n",
       "      <th>created_at</th>\n",
       "      <th>phone_type</th>\n",
       "      <th>bureau_score</th>\n",
       "      <th>spend_desire</th>\n",
       "      <th>random_noise</th>\n",
       "      <th>monthly_spend</th>\n",
       "      <th>month_date</th>\n",
       "      <th>avg_last_2_months_spend</th>\n",
       "      <th>target</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>17</td>\n",
       "      <td>5015.685403</td>\n",
       "      <td>2018-05-14</td>\n",
       "      <td>samsung</td>\n",
       "      <td>NaN</td>\n",
       "      <td>759.448286</td>\n",
       "      <td>929.006227</td>\n",
       "      <td>3452.226271</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>NaN</td>\n",
       "      <td>3516.757404</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>18</td>\n",
       "      <td>5015.685403</td>\n",
       "      <td>2018-05-14</td>\n",
       "      <td>samsung</td>\n",
       "      <td>522.001029</td>\n",
       "      <td>759.448286</td>\n",
       "      <td>1103.297890</td>\n",
       "      <td>3880.094787</td>\n",
       "      <td>2018-07-13</td>\n",
       "      <td>3666.160529</td>\n",
       "      <td>3228.706602</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>19</td>\n",
       "      <td>5015.685403</td>\n",
       "      <td>2018-05-14</td>\n",
       "      <td>samsung</td>\n",
       "      <td>381.931691</td>\n",
       "      <td>759.448286</td>\n",
       "      <td>970.967715</td>\n",
       "      <td>3153.420020</td>\n",
       "      <td>2018-08-13</td>\n",
       "      <td>3516.757404</td>\n",
       "      <td>3357.249359</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>20</td>\n",
       "      <td>5015.685403</td>\n",
       "      <td>2018-05-14</td>\n",
       "      <td>samsung</td>\n",
       "      <td>316.660045</td>\n",
       "      <td>759.448286</td>\n",
       "      <td>960.317678</td>\n",
       "      <td>3303.993183</td>\n",
       "      <td>2018-09-13</td>\n",
       "      <td>3228.706602</td>\n",
       "      <td>3358.220025</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>21</td>\n",
       "      <td>5015.685403</td>\n",
       "      <td>2018-05-14</td>\n",
       "      <td>samsung</td>\n",
       "      <td>319.226632</td>\n",
       "      <td>759.448286</td>\n",
       "      <td>940.807237</td>\n",
       "      <td>3410.505535</td>\n",
       "      <td>2018-10-14</td>\n",
       "      <td>3357.249359</td>\n",
       "      <td>3235.362465</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   id  month       income  created_at phone_type  bureau_score  spend_desire  \\\n",
       "0   0     17  5015.685403  2018-05-14    samsung           NaN    759.448286   \n",
       "1   0     18  5015.685403  2018-05-14    samsung    522.001029    759.448286   \n",
       "2   0     19  5015.685403  2018-05-14    samsung    381.931691    759.448286   \n",
       "3   0     20  5015.685403  2018-05-14    samsung    316.660045    759.448286   \n",
       "4   0     21  5015.685403  2018-05-14    samsung    319.226632    759.448286   \n",
       "\n",
       "   random_noise  monthly_spend month_date  avg_last_2_months_spend  \\\n",
       "0    929.006227    3452.226271 2018-06-12                      NaN   \n",
       "1   1103.297890    3880.094787 2018-07-13              3666.160529   \n",
       "2    970.967715    3153.420020 2018-08-13              3516.757404   \n",
       "3    960.317678    3303.993183 2018-09-13              3228.706602   \n",
       "4    940.807237    3410.505535 2018-10-14              3357.249359   \n",
       "\n",
       "        target  \n",
       "0  3516.757404  \n",
       "1  3228.706602  \n",
       "2  3357.249359  \n",
       "3  3358.220025  \n",
       "4  3235.362465  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:18:14.525863Z",
     "start_time": "2019-09-26T20:18:14.438602Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "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>count</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>min</th>\n",
       "      <th>25%</th>\n",
       "      <th>50%</th>\n",
       "      <th>75%</th>\n",
       "      <th>max</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>id</th>\n",
       "      <td>121471.0</td>\n",
       "      <td>5020.204798</td>\n",
       "      <td>2.882829e+03</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>2532.000000</td>\n",
       "      <td>5012.000000</td>\n",
       "      <td>7531.000000</td>\n",
       "      <td>9.999000e+03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>month</th>\n",
       "      <td>121471.0</td>\n",
       "      <td>15.629920</td>\n",
       "      <td>5.541509e+00</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>12.000000</td>\n",
       "      <td>17.000000</td>\n",
       "      <td>20.000000</td>\n",
       "      <td>2.300000e+01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>income</th>\n",
       "      <td>121471.0</td>\n",
       "      <td>505248.655911</td>\n",
       "      <td>2.179283e+06</td>\n",
       "      <td>302.532705</td>\n",
       "      <td>3758.661316</td>\n",
       "      <td>5179.311664</td>\n",
       "      <td>6627.112964</td>\n",
       "      <td>9.999999e+06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>bureau_score</th>\n",
       "      <td>109433.0</td>\n",
       "      <td>295.955427</td>\n",
       "      <td>1.346843e+02</td>\n",
       "      <td>0.005788</td>\n",
       "      <td>198.558112</td>\n",
       "      <td>293.211936</td>\n",
       "      <td>388.348340</td>\n",
       "      <td>9.665136e+02</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>spend_desire</th>\n",
       "      <td>121471.0</td>\n",
       "      <td>497.221982</td>\n",
       "      <td>2.024356e+02</td>\n",
       "      <td>-356.866864</td>\n",
       "      <td>360.737656</td>\n",
       "      <td>497.799260</td>\n",
       "      <td>635.393654</td>\n",
       "      <td>1.260928e+03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>random_noise</th>\n",
       "      <td>121471.0</td>\n",
       "      <td>999.910829</td>\n",
       "      <td>1.000080e+02</td>\n",
       "      <td>514.929020</td>\n",
       "      <td>932.140569</td>\n",
       "      <td>999.841128</td>\n",
       "      <td>1067.019196</td>\n",
       "      <td>1.428127e+03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>monthly_spend</th>\n",
       "      <td>121471.0</td>\n",
       "      <td>2643.700884</td>\n",
       "      <td>6.920308e+02</td>\n",
       "      <td>170.815303</td>\n",
       "      <td>2192.303084</td>\n",
       "      <td>2554.895551</td>\n",
       "      <td>2961.684065</td>\n",
       "      <td>6.537560e+03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>avg_last_2_months_spend</th>\n",
       "      <td>111471.0</td>\n",
       "      <td>2643.832147</td>\n",
       "      <td>5.860818e+02</td>\n",
       "      <td>188.307429</td>\n",
       "      <td>2238.787433</td>\n",
       "      <td>2585.316456</td>\n",
       "      <td>2984.386894</td>\n",
       "      <td>5.782090e+03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>target</th>\n",
       "      <td>101871.0</td>\n",
       "      <td>2639.039170</td>\n",
       "      <td>5.853034e+02</td>\n",
       "      <td>412.076860</td>\n",
       "      <td>2235.043343</td>\n",
       "      <td>2580.730663</td>\n",
       "      <td>2978.910581</td>\n",
       "      <td>5.782090e+03</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                            count           mean           std         min  \\\n",
       "id                       121471.0    5020.204798  2.882829e+03    0.000000   \n",
       "month                    121471.0      15.629920  5.541509e+00    1.000000   \n",
       "income                   121471.0  505248.655911  2.179283e+06  302.532705   \n",
       "bureau_score             109433.0     295.955427  1.346843e+02    0.005788   \n",
       "spend_desire             121471.0     497.221982  2.024356e+02 -356.866864   \n",
       "random_noise             121471.0     999.910829  1.000080e+02  514.929020   \n",
       "monthly_spend            121471.0    2643.700884  6.920308e+02  170.815303   \n",
       "avg_last_2_months_spend  111471.0    2643.832147  5.860818e+02  188.307429   \n",
       "target                   101871.0    2639.039170  5.853034e+02  412.076860   \n",
       "\n",
       "                                 25%          50%          75%           max  \n",
       "id                       2532.000000  5012.000000  7531.000000  9.999000e+03  \n",
       "month                      12.000000    17.000000    20.000000  2.300000e+01  \n",
       "income                   3758.661316  5179.311664  6627.112964  9.999999e+06  \n",
       "bureau_score              198.558112   293.211936   388.348340  9.665136e+02  \n",
       "spend_desire              360.737656   497.799260   635.393654  1.260928e+03  \n",
       "random_noise              932.140569   999.841128  1067.019196  1.428127e+03  \n",
       "monthly_spend            2192.303084  2554.895551  2961.684065  6.537560e+03  \n",
       "avg_last_2_months_spend  2238.787433  2585.316456  2984.386894  5.782090e+03  \n",
       "target                   2235.043343  2580.730663  2978.910581  5.782090e+03  "
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.describe().T"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:18:17.033191Z",
     "start_time": "2019-09-26T20:18:17.030190Z"
    },
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [],
   "source": [
    "features = [\"income\", \"bureau_score\", \"spend_desire\", \"random_noise\", \"monthly_spend\", \"avg_last_2_months_spend\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:18:21.117236Z",
     "start_time": "2019-09-26T20:18:21.078410Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "id                             0\n",
       "month                          0\n",
       "income                         0\n",
       "created_at                     0\n",
       "phone_type                     0\n",
       "bureau_score               12038\n",
       "spend_desire                   0\n",
       "random_noise                   0\n",
       "monthly_spend                  0\n",
       "month_date                     0\n",
       "avg_last_2_months_spend    10000\n",
       "target                     19600\n",
       "dtype: int64"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.isna().sum()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## Features and Target:\n",
    "$ Month (M) \\sim [1, 24]  $\n",
    "\n",
    "$ Income (I) \\sim \\mathcal{N}(5000, 2000) E [300, 20000] $\n",
    "\n",
    "$ Phone Type (P) \\sim [\"samsung\", \"motorola\", \"iphone\", \"lg\"] $\n",
    "\n",
    "$ Bureau Score (B) \\sim \\mathcal{N}(\\frac{500}{(Month ^{0.1}}, 200) E [0, 1000] $\n",
    "\n",
    "$ Spend Desire (W) \\sim \\mathcal{N}(500, 200) $\n",
    "\n",
    "$ Random Noise (R) \\sim \\mathcal{N}(1000, 100) $\n",
    "\n",
    "$ Monthly Spend \\sim {max}(0, (S * I + I ^ 2 + P * W ^ 2 + P * B + \\mathcal{N}(1, 0.3)) * \\mathcal{N}(2000, 1000)) $\n",
    "\n",
    "$ Avg Last 2 Months Spend \\sim (Spend(m) + Spend(m-1)) / 2 $\n",
    "\n",
    "$ Target \\sim (Spend(m + 1) + Spend(m + 2)) / 2 $\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:33:19.160207Z",
     "start_time": "2019-09-26T20:33:18.964129Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Amount of customers by month')"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "plt.plot(sorted(df.month.unique()), df.groupby(\"month\").agg({\"id\": \"count\"}))\n",
    "plt.title(\"Amount of customers by month\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:33:22.408901Z",
     "start_time": "2019-09-26T20:33:21.774585Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, axes = plt.subplots(2, 2, figsize=(8,8))\n",
    "axes[0, 0].hist(df.bureau_score, range(0, 1000, 50))\n",
    "axes[1, 0].hist(df.income, range(300, 20000, 500))\n",
    "axes[0, 1].hist(df.spend_desire)\n",
    "axes[1, 1].hist(df.monthly_spend, range(0, 10000, 500))\n",
    "\n",
    "titles = [\"bureau_histogram\", \"income_histogram\", \"spend_desire_histogram\", \"monthly_spend_histogram\"]\n",
    "axes[0, 0].set_title(titles[0])\n",
    "axes[1, 0].set_title(titles[1])\n",
    "axes[0, 1].set_title(titles[2])\n",
    "axes[1, 1].set_title(titles[3])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:33:34.377343Z",
     "start_time": "2019-09-26T20:33:24.855505Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, axes = plt.subplots(2, 2, figsize=(10,10))\n",
    "sns.pointplot(x=\"month\", y=\"bureau_score\", data=df, ax=axes[0, 0])\n",
    "sns.pointplot(x=\"month\", y=\"income\", data=df, ax=axes[1, 0])\n",
    "sns.pointplot(x=\"month\", y=\"spend_desire\", data=df, ax=axes[0, 1])\n",
    "sns.pointplot(x=\"month\", y=\"monthly_spend\", data=df, ax=axes[1, 1])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## Target Analysis"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:33:34.414327Z",
     "start_time": "2019-09-26T20:33:34.379476Z"
    },
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "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>null_count_by_month</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>month_date</th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2017-02-01</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2017-03-04</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2017-04-04</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2017-05-05</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2017-06-05</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2017-07-06</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2017-08-06</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2017-09-06</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2017-10-07</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2017-11-07</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2017-12-08</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-01-08</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-02-08</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-03-11</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-04-11</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-05-12</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-06-12</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-07-13</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-08-13</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-09-13</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-10-14</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-11-14</th>\n",
       "      <td>9600</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-12-15</th>\n",
       "      <td>10000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            null_count_by_month\n",
       "month_date                     \n",
       "2017-02-01                    0\n",
       "2017-03-04                    0\n",
       "2017-04-04                    0\n",
       "2017-05-05                    0\n",
       "2017-06-05                    0\n",
       "2017-07-06                    0\n",
       "2017-08-06                    0\n",
       "2017-09-06                    0\n",
       "2017-10-07                    0\n",
       "2017-11-07                    0\n",
       "2017-12-08                    0\n",
       "2018-01-08                    0\n",
       "2018-02-08                    0\n",
       "2018-03-11                    0\n",
       "2018-04-11                    0\n",
       "2018-05-12                    0\n",
       "2018-06-12                    0\n",
       "2018-07-13                    0\n",
       "2018-08-13                    0\n",
       "2018-09-13                    0\n",
       "2018-10-14                    0\n",
       "2018-11-14                 9600\n",
       "2018-12-15                10000"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.DataFrame(df.groupby(\"month_date\").apply(lambda x: x.target.isna().sum()), columns=[\"null_count_by_month\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:33:36.514225Z",
     "start_time": "2019-09-26T20:33:34.417090Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x10bce9748>"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "sns.pointplot(x=\"month\", y=\"target\", data=df)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# Getting started with fklearn: Creating a simple model\n",
    "\n",
    "* <font size=\"4\"> FKlearn is focused on creating production ready models but is flexible to be used at research and learning tasks</font>\n",
    "\n",
    "\n",
    "* <font size=\"4\"> Tradional steps of training a model like: Train-Test Splitting, Training and Validating are all implemented on fklearn </font>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## Spliting the dataset into train and holdout\n",
    "\n",
    "* <font size=\"4\"> On real problems we want to validate our model both in the same period and also on a different period to guarantee our model is a good predictor for the future </font>\n",
    "\n",
    "\n",
    "* <font size=\"4\"> On this case of customers spend we also would like to validate the performance of the model on a different set of customers to check for overfitting in the population </font>\n",
    "<img src=\"imgs/Splitting_Data.png\">"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:49:37.576431Z",
     "start_time": "2019-09-26T20:49:37.529755Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "from fklearn.preprocessing.splitting import space_time_split_dataset\n",
    "\n",
    "train_set, intime_outspace_hdout, outime_inspace_hdout, outime_outspace_hdout = \\\n",
    "                                        space_time_split_dataset(df,\n",
    "                                                 train_start_date=\"2017-02-01\",\n",
    "                                                 train_end_date=\"2018-07-01\",\n",
    "                                                 holdout_end_date=\"2018-10-01\",\n",
    "                                                 split_seed=42,\n",
    "                                                 space_holdout_percentage=0.2,\n",
    "                                                 space_column=\"id\",\n",
    "                                                 time_column=\"month_date\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:49:38.861828Z",
     "start_time": "2019-09-26T20:49:38.857455Z"
    },
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((54206, 12), (13421, 12), (25055, 12), (4488, 12))"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(train_set.shape, \n",
    " intime_outspace_hdout.shape, \n",
    " outime_inspace_hdout.shape,\n",
    " outime_outspace_hdout.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## Training:\n",
    "<font size=\"4\"> *On the training process we want:* </font>\n",
    "\n",
    "* <font size=\"4\">  **Cap features** that have unexpected values </font>\n",
    "\n",
    "\n",
    "* <font size=\"4\"> **Encode categorical features** </font>\n",
    "\n",
    "\n",
    "* <font size=\"4\">  **Fill missing values** </font>\n",
    "\n",
    "\n",
    "* <font size=\"4\">  **Train our model** </font>\n",
    "\n",
    "\n",
    "* <font size=\"4\"> We want to apply this **same transformations when scoring our model**, similarly to the fit and transform concept from sklearn </font>\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-04-24T16:17:39.071198Z",
     "start_time": "2019-04-24T16:17:39.067285Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "\n",
    "<img src=\"imgs/Pipeline.png\">"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n",
    "<img src=\"imgs/partial-functions.png\">"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:49:44.466374Z",
     "start_time": "2019-09-26T20:49:43.895409Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "from fklearn.training.transformation import capper, prediction_ranger, label_categorizer\n",
    "from fklearn.training.imputation import imputer\n",
    "from fklearn.training.regression import lgbm_regression_learner\n",
    "capper_fn = capper(columns_to_cap=[\"income\"], precomputed_caps={\"income\": 20000.0})\n",
    "ranger_fn = prediction_ranger(prediction_min=0.0, prediction_max=100000.0, prediction_column=\"prediction\")\n",
    "label_fn = label_categorizer(columns_to_categorize=[\"phone_type\"])\n",
    "imputer_fn = imputer(columns_to_impute=[\"bureau_score\"], impute_strategy=\"median\")\n",
    "regression_fn = lgbm_regression_learner(features=features, target=\"target\", learning_rate=0.1, num_estimators=200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:49:44.749046Z",
     "start_time": "2019-09-26T20:49:44.742850Z"
    },
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [],
   "source": [
    "from fklearn.training.pipeline import build_pipeline\n",
    "train_fn = build_pipeline(label_fn, capper_fn, imputer_fn, regression_fn, ranger_fn)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:49:46.059635Z",
     "start_time": "2019-09-26T20:49:45.463989Z"
    },
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/henriquelopes/miniconda3/lib/python3.6/site-packages/sklearn/utils/deprecation.py:66: DeprecationWarning: Class Imputer is deprecated; Imputer was deprecated in version 0.20 and will be removed in 0.22. Import impute.SimpleImputer from sklearn instead.\n",
      "  warnings.warn(msg, category=DeprecationWarning)\n"
     ]
    }
   ],
   "source": [
    "predict_fn, scored_train_set, train_logs = train_fn(train_set)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:49:46.084073Z",
     "start_time": "2019-09-26T20:49:46.062419Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "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>id</th>\n",
       "      <th>month</th>\n",
       "      <th>income</th>\n",
       "      <th>created_at</th>\n",
       "      <th>phone_type</th>\n",
       "      <th>bureau_score</th>\n",
       "      <th>spend_desire</th>\n",
       "      <th>random_noise</th>\n",
       "      <th>monthly_spend</th>\n",
       "      <th>month_date</th>\n",
       "      <th>avg_last_2_months_spend</th>\n",
       "      <th>target</th>\n",
       "      <th>prediction</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>2</td>\n",
       "      <td>3</td>\n",
       "      <td>2424.105189</td>\n",
       "      <td>2017-03-28</td>\n",
       "      <td>0</td>\n",
       "      <td>611.650913</td>\n",
       "      <td>641.98234</td>\n",
       "      <td>858.263250</td>\n",
       "      <td>3059.915274</td>\n",
       "      <td>2017-04-04</td>\n",
       "      <td>NaN</td>\n",
       "      <td>3492.075105</td>\n",
       "      <td>2886.165513</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>2</td>\n",
       "      <td>4</td>\n",
       "      <td>20000.000000</td>\n",
       "      <td>2017-03-28</td>\n",
       "      <td>0</td>\n",
       "      <td>310.829209</td>\n",
       "      <td>641.98234</td>\n",
       "      <td>984.614824</td>\n",
       "      <td>4447.520879</td>\n",
       "      <td>2017-05-05</td>\n",
       "      <td>3753.718077</td>\n",
       "      <td>2425.698944</td>\n",
       "      <td>2788.816923</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>2</td>\n",
       "      <td>5</td>\n",
       "      <td>2424.105189</td>\n",
       "      <td>2017-03-28</td>\n",
       "      <td>0</td>\n",
       "      <td>310.829209</td>\n",
       "      <td>641.98234</td>\n",
       "      <td>1004.881140</td>\n",
       "      <td>2536.629332</td>\n",
       "      <td>2017-06-05</td>\n",
       "      <td>3492.075105</td>\n",
       "      <td>2256.332465</td>\n",
       "      <td>2783.500567</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>2</td>\n",
       "      <td>6</td>\n",
       "      <td>2424.105189</td>\n",
       "      <td>2017-03-28</td>\n",
       "      <td>0</td>\n",
       "      <td>156.248181</td>\n",
       "      <td>641.98234</td>\n",
       "      <td>960.071795</td>\n",
       "      <td>2314.768557</td>\n",
       "      <td>2017-07-06</td>\n",
       "      <td>2425.698944</td>\n",
       "      <td>3476.051462</td>\n",
       "      <td>2758.309663</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>2</td>\n",
       "      <td>7</td>\n",
       "      <td>2424.105189</td>\n",
       "      <td>2017-03-28</td>\n",
       "      <td>0</td>\n",
       "      <td>167.091261</td>\n",
       "      <td>641.98234</td>\n",
       "      <td>991.301503</td>\n",
       "      <td>2197.896373</td>\n",
       "      <td>2017-08-06</td>\n",
       "      <td>2256.332465</td>\n",
       "      <td>3648.279899</td>\n",
       "      <td>2789.522977</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    id  month        income  created_at  phone_type  bureau_score  \\\n",
       "9    2      3   2424.105189  2017-03-28           0    611.650913   \n",
       "10   2      4  20000.000000  2017-03-28           0    310.829209   \n",
       "11   2      5   2424.105189  2017-03-28           0    310.829209   \n",
       "12   2      6   2424.105189  2017-03-28           0    156.248181   \n",
       "13   2      7   2424.105189  2017-03-28           0    167.091261   \n",
       "\n",
       "    spend_desire  random_noise  monthly_spend month_date  \\\n",
       "9      641.98234    858.263250    3059.915274 2017-04-04   \n",
       "10     641.98234    984.614824    4447.520879 2017-05-05   \n",
       "11     641.98234   1004.881140    2536.629332 2017-06-05   \n",
       "12     641.98234    960.071795    2314.768557 2017-07-06   \n",
       "13     641.98234    991.301503    2197.896373 2017-08-06   \n",
       "\n",
       "    avg_last_2_months_spend       target   prediction  \n",
       "9                       NaN  3492.075105  2886.165513  \n",
       "10              3753.718077  2425.698944  2788.816923  \n",
       "11              3492.075105  2256.332465  2783.500567  \n",
       "12              2425.698944  3476.051462  2758.309663  \n",
       "13              2256.332465  3648.279899  2789.522977  "
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scored_train_set.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:49:48.433435Z",
     "start_time": "2019-09-26T20:49:48.424113Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "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>importance</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>avg_last_2_months_spend</th>\n",
       "      <td>1149</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>spend_desire</th>\n",
       "      <td>1129</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>bureau_score</th>\n",
       "      <td>987</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>income</th>\n",
       "      <td>967</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>monthly_spend</th>\n",
       "      <td>960</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>random_noise</th>\n",
       "      <td>808</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                         importance\n",
       "avg_last_2_months_spend        1149\n",
       "spend_desire                   1129\n",
       "bureau_score                    987\n",
       "income                          967\n",
       "monthly_spend                   960\n",
       "random_noise                    808"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.options.display.max_rows = 120\n",
    "pd.DataFrame(train_logs['lgbm_regression_learner']['feature_importance'], \n",
    "             index=[\"importance\"]).T.sort_values(\"importance\", ascending=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:49:53.299682Z",
     "start_time": "2019-09-26T20:49:53.115017Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "scored_outime_outspace_hdout = predict_fn(outime_outspace_hdout)\n",
    "scored_intime_outspace_hdout = predict_fn(intime_outspace_hdout)\n",
    "scored_outime_inspace_hdout = predict_fn(outime_inspace_hdout)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:49:54.013179Z",
     "start_time": "2019-09-26T20:49:53.938932Z"
    },
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "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>id</th>\n",
       "      <th>month_date</th>\n",
       "      <th>income</th>\n",
       "      <th>bureau_score</th>\n",
       "      <th>target</th>\n",
       "      <th>prediction</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>2018-07-13</td>\n",
       "      <td>5015.685403</td>\n",
       "      <td>522.001029</td>\n",
       "      <td>3228.706602</td>\n",
       "      <td>3451.870684</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>2018-08-13</td>\n",
       "      <td>5015.685403</td>\n",
       "      <td>381.931691</td>\n",
       "      <td>3357.249359</td>\n",
       "      <td>3210.763017</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>2018-09-13</td>\n",
       "      <td>5015.685403</td>\n",
       "      <td>316.660045</td>\n",
       "      <td>3358.220025</td>\n",
       "      <td>3385.630512</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>148</th>\n",
       "      <td>13</td>\n",
       "      <td>2018-07-13</td>\n",
       "      <td>20000.000000</td>\n",
       "      <td>330.107289</td>\n",
       "      <td>2107.578901</td>\n",
       "      <td>2398.327320</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>149</th>\n",
       "      <td>13</td>\n",
       "      <td>2018-08-13</td>\n",
       "      <td>1731.884844</td>\n",
       "      <td>260.790931</td>\n",
       "      <td>2091.307307</td>\n",
       "      <td>2328.283377</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     id month_date        income  bureau_score       target   prediction\n",
       "1     0 2018-07-13   5015.685403    522.001029  3228.706602  3451.870684\n",
       "2     0 2018-08-13   5015.685403    381.931691  3357.249359  3210.763017\n",
       "3     0 2018-09-13   5015.685403    316.660045  3358.220025  3385.630512\n",
       "148  13 2018-07-13  20000.000000    330.107289  2107.578901  2398.327320\n",
       "149  13 2018-08-13   1731.884844    260.790931  2091.307307  2328.283377"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scored_outime_outspace_hdout[[\"id\", \"month_date\", \"income\", \"bureau_score\", \"target\", \"prediction\"]].head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## Evaluation:\n",
    "* <font size=\"4\"> We want to evaluate the performance of our model using multiple metrics for each dataframe (out_of_time, out_of_space, out_of_time_and_space) </font>\n",
    "\n",
    "\n",
    "<img src=\"imgs/EvalFn.png\">"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:49:55.409119Z",
     "start_time": "2019-09-26T20:49:55.402097Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "from fklearn.validation.evaluators import r2_evaluator, spearman_evaluator, combined_evaluators"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:49:55.836022Z",
     "start_time": "2019-09-26T20:49:55.832101Z"
    },
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [],
   "source": [
    "r2_eval_fn = r2_evaluator(prediction_column=\"prediction\", target_column=\"target\")\n",
    "spearman_eval_fn = spearman_evaluator(prediction_column=\"prediction\", target_column=\"target\")\n",
    "eval_fn = combined_evaluators(evaluators=[r2_eval_fn, spearman_eval_fn])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:49:56.333094Z",
     "start_time": "2019-09-26T20:49:56.292794Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "outime_outspace_hdout_logs = eval_fn(scored_outime_outspace_hdout)\n",
    "intime_outspace_hdout_logs = eval_fn(scored_intime_outspace_hdout)\n",
    "outime_inspace_hdout_logs = eval_fn(scored_outime_inspace_hdout)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:49:56.773191Z",
     "start_time": "2019-09-26T20:49:56.768896Z"
    },
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'out_of_time': {'r2_evaluator__target': 0.5956009555286967,\n",
       "  'spearman_evaluator__target': 0.7892796752342339},\n",
       " 'in_time': {'r2_evaluator__target': 0.588576925917941,\n",
       "  'spearman_evaluator__target': 0.787194941139121},\n",
       " 'out_of_time_and_space': {'r2_evaluator__target': 0.5749628505861282,\n",
       "  'spearman_evaluator__target': 0.7790056568693072}}"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "{\"out_of_time\": outime_outspace_hdout_logs, \n",
    " \"in_time\": intime_outspace_hdout_logs,\n",
    "  \"out_of_time_and_space\": outime_inspace_hdout_logs}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## Extractors\n",
    "* <font size=\"4\"> We want to transform this logs into dataframes that we can visualize better </font>\n",
    "\n",
    "<img src=\"imgs/Extractor.png\" style=\"width: 500px;\">"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:49:58.042135Z",
     "start_time": "2019-09-26T20:49:58.033949Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "from fklearn.metrics.pd_extractors import evaluator_extractor, combined_evaluator_extractor, extract\n",
    "r2_extractor = evaluator_extractor(evaluator_name=\"r2_evaluator__target\")\n",
    "spearman_extractor = evaluator_extractor(evaluator_name=\"spearman_evaluator__target\")\n",
    "full_extractor = combined_evaluator_extractor(base_extractors=[r2_extractor, spearman_extractor])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:49:58.519777Z",
     "start_time": "2019-09-26T20:49:58.504037Z"
    },
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "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>r2_evaluator__target</th>\n",
       "      <th>spearman_evaluator__target</th>\n",
       "      <th>part</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.595601</td>\n",
       "      <td>0.789280</td>\n",
       "      <td>out_of_time</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.588577</td>\n",
       "      <td>0.787195</td>\n",
       "      <td>in_time_out_of_space</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.574963</td>\n",
       "      <td>0.779006</td>\n",
       "      <td>out_of_time_and_space</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   r2_evaluator__target  spearman_evaluator__target                   part\n",
       "0              0.595601                    0.789280            out_of_time\n",
       "0              0.588577                    0.787195   in_time_out_of_space\n",
       "0              0.574963                    0.779006  out_of_time_and_space"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.concat(\n",
    "    [full_extractor(outime_outspace_hdout_logs).assign(part=\"out_of_time\"),\n",
    "    full_extractor(intime_outspace_hdout_logs).assign(part=\"in_time_out_of_space\"),\n",
    "    full_extractor(outime_inspace_hdout_logs).assign(part=\"out_of_time_and_space\")])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# Learning Curves:\n",
    "\n",
    "<img src=\"imgs/LearningCurves.png\">\n",
    "\n",
    "<font size=\"5\"> We can reuse our previously defined training function and evaluators! </font>\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## Could we improve if we had more data?? AKA Spatial Learning Curve\n",
    "\n",
    "* <font size=\"4\"> If we had more data we could use that for training and see how much the model improves </font>\n",
    "\n",
    "\n",
    "* <font size=\"4\"> To estimate that we can train with subsamples of the full training set and check how performance changes </font>\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "<img src=\"imgs/SpatialLearningCurves.png\">"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-26T11:16:10.506298Z",
     "start_time": "2019-06-26T11:16:10.468783Z"
    }
   },
   "source": [
    "<img src=\"imgs/validation-explanation.png\">"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:50:03.265747Z",
     "start_time": "2019-09-26T20:50:03.257797Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "from fklearn.validation.splitters import spatial_learning_curve_splitter\n",
    "\n",
    "split_fn = spatial_learning_curve_splitter(space_column=\"id\", \n",
    "                                           time_column=\"month_date\",\n",
    "                                           training_limit=\"2018-04-01\", \n",
    "                                           train_percentages=[0.1, 0.2, 0.4, 0.6, 0.8, 1.0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:50:12.703263Z",
     "start_time": "2019-09-26T20:50:03.719655Z"
    },
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/henriquelopes/miniconda3/lib/python3.6/site-packages/sklearn/utils/deprecation.py:66: DeprecationWarning: Class Imputer is deprecated; Imputer was deprecated in version 0.20 and will be removed in 0.22. Import impute.SimpleImputer from sklearn instead.\n",
      "  warnings.warn(msg, category=DeprecationWarning)\n",
      "/Users/henriquelopes/miniconda3/lib/python3.6/site-packages/sklearn/utils/deprecation.py:66: DeprecationWarning: Class Imputer is deprecated; Imputer was deprecated in version 0.20 and will be removed in 0.22. Import impute.SimpleImputer from sklearn instead.\n",
      "  warnings.warn(msg, category=DeprecationWarning)\n",
      "/Users/henriquelopes/miniconda3/lib/python3.6/site-packages/sklearn/utils/deprecation.py:66: DeprecationWarning: Class Imputer is deprecated; Imputer was deprecated in version 0.20 and will be removed in 0.22. Import impute.SimpleImputer from sklearn instead.\n",
      "  warnings.warn(msg, category=DeprecationWarning)\n",
      "/Users/henriquelopes/miniconda3/lib/python3.6/site-packages/sklearn/utils/deprecation.py:66: DeprecationWarning: Class Imputer is deprecated; Imputer was deprecated in version 0.20 and will be removed in 0.22. Import impute.SimpleImputer from sklearn instead.\n",
      "  warnings.warn(msg, category=DeprecationWarning)\n",
      "/Users/henriquelopes/miniconda3/lib/python3.6/site-packages/sklearn/utils/deprecation.py:66: DeprecationWarning: Class Imputer is deprecated; Imputer was deprecated in version 0.20 and will be removed in 0.22. Import impute.SimpleImputer from sklearn instead.\n",
      "  warnings.warn(msg, category=DeprecationWarning)\n",
      "/Users/henriquelopes/miniconda3/lib/python3.6/site-packages/sklearn/utils/deprecation.py:66: DeprecationWarning: Class Imputer is deprecated; Imputer was deprecated in version 0.20 and will be removed in 0.22. Import impute.SimpleImputer from sklearn instead.\n",
      "  warnings.warn(msg, category=DeprecationWarning)\n"
     ]
    }
   ],
   "source": [
    "from fklearn.validation.validator import parallel_validator\n",
    "\n",
    "spatial_learning_curve_logs = parallel_validator(train_set, split_fn, train_fn, eval_fn, n_jobs=8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:50:12.742892Z",
     "start_time": "2019-09-26T20:50:12.705829Z"
    },
    "scrolled": true,
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'train_log': [{'label_categorizer': {'transformed_column': ['phone_type'],\n",
       "    'replace_unseen': nan,\n",
       "    'running_time': '0.022 s'},\n",
       "   'capper': {'caps': {'income': 20000.0},\n",
       "    'transformed_column': ['income'],\n",
       "    'precomputed_caps': {'income': 20000.0},\n",
       "    'running_time': '0.076 s'},\n",
       "   'imputer': {'impute_strategy': 'median',\n",
       "    'columns_to_impute': ['bureau_score'],\n",
       "    'training_proportion_of_nulls': {'bureau_score': 0.09924550203134068},\n",
       "    'statistics': array([325.16258753]),\n",
       "    'running_time': '0.038 s'},\n",
       "   'lgbm_regression_learner': {'features': ['income',\n",
       "     'bureau_score',\n",
       "     'spend_desire',\n",
       "     'random_noise',\n",
       "     'monthly_spend',\n",
       "     'avg_last_2_months_spend'],\n",
       "    'target': 'target',\n",
       "    'prediction_column': 'prediction',\n",
       "    'package': 'lightgbm',\n",
       "    'package_version': '2.2.3',\n",
       "    'parameters': {'eta': 0.1,\n",
       "     'objective': 'regression',\n",
       "     'num_estimators': 200},\n",
       "    'feature_importance': {'income': 979,\n",
       "     'bureau_score': 1008,\n",
       "     'spend_desire': 1048,\n",
       "     'random_noise': 958,\n",
       "     'monthly_spend': 844,\n",
       "     'avg_last_2_months_spend': 1163},\n",
       "    'training_samples': 5169,\n",
       "    'running_time': '7.715 s'},\n",
       "   'object': <lightgbm.basic.Booster at 0x1a2263c5f8>,\n",
       "   'prediction_ranger': {'prediction_min': 0.0,\n",
       "    'prediction_max': 100000.0,\n",
       "    'transformed_column': ['prediction']},\n",
       "   '__fkml__': {'pipeline': ['label_categorizer',\n",
       "     'capper',\n",
       "     'imputer',\n",
       "     'lgbm_regression_learner',\n",
       "     'prediction_ranger'],\n",
       "    'output_columns': ['id',\n",
       "     'month',\n",
       "     'income',\n",
       "     'created_at',\n",
       "     'phone_type',\n",
       "     'bureau_score',\n",
       "     'spend_desire',\n",
       "     'random_noise',\n",
       "     'monthly_spend',\n",
       "     'month_date',\n",
       "     'avg_last_2_months_spend',\n",
       "     'target',\n",
       "     'prediction'],\n",
       "    'features': ['id',\n",
       "     'month',\n",
       "     'income',\n",
       "     'created_at',\n",
       "     'phone_type',\n",
       "     'bureau_score',\n",
       "     'spend_desire',\n",
       "     'random_noise',\n",
       "     'monthly_spend',\n",
       "     'month_date',\n",
       "     'avg_last_2_months_spend',\n",
       "     'target'],\n",
       "    'learners': {'label_categorizer': {'fn': <function fklearn.training.transformation.label_categorizer.<locals>.p(new_df:pandas.core.frame.DataFrame) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'label_categorizer': {'transformed_column': ['phone_type'],\n",
       "        'replace_unseen': nan,\n",
       "        'running_time': '0.022 s'}}},\n",
       "     'capper': {'fn': <function fklearn.training.transformation.capper.<locals>.p(new_data_set:pandas.core.frame.DataFrame) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'capper': {'caps': {'income': 20000.0},\n",
       "        'transformed_column': ['income'],\n",
       "        'precomputed_caps': {'income': 20000.0},\n",
       "        'running_time': '0.076 s'}}},\n",
       "     'imputer': {'fn': <function fklearn.training.imputation.imputer.<locals>.p(new_data_set:pandas.core.frame.DataFrame) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'imputer': {'impute_strategy': 'median',\n",
       "        'columns_to_impute': ['bureau_score'],\n",
       "        'training_proportion_of_nulls': {'bureau_score': 0.09924550203134068},\n",
       "        'statistics': array([325.16258753]),\n",
       "        'running_time': '0.038 s'}}},\n",
       "     'lgbm_regression_learner': {'fn': <function fklearn.training.regression.lgbm_regression_learner.<locals>.p(new_df:pandas.core.frame.DataFrame, apply_shap:bool=False) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'lgbm_regression_learner': {'features': ['income',\n",
       "         'bureau_score',\n",
       "         'spend_desire',\n",
       "         'random_noise',\n",
       "         'monthly_spend',\n",
       "         'avg_last_2_months_spend'],\n",
       "        'target': 'target',\n",
       "        'prediction_column': 'prediction',\n",
       "        'package': 'lightgbm',\n",
       "        'package_version': '2.2.3',\n",
       "        'parameters': {'eta': 0.1,\n",
       "         'objective': 'regression',\n",
       "         'num_estimators': 200},\n",
       "        'feature_importance': {'income': 979,\n",
       "         'bureau_score': 1008,\n",
       "         'spend_desire': 1048,\n",
       "         'random_noise': 958,\n",
       "         'monthly_spend': 844,\n",
       "         'avg_last_2_months_spend': 1163},\n",
       "        'training_samples': 5169,\n",
       "        'running_time': '7.715 s'},\n",
       "       'object': <lightgbm.basic.Booster at 0x1a2263c5f8>}},\n",
       "     'prediction_ranger': {'fn': <function fklearn.training.transformation.prediction_ranger.<locals>.p(new_df:pandas.core.frame.DataFrame) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'prediction_ranger': {'prediction_min': 0.0,\n",
       "        'prediction_max': 100000.0,\n",
       "        'transformed_column': ['prediction']}}}}}},\n",
       "  {'label_categorizer': {'transformed_column': ['phone_type'],\n",
       "    'replace_unseen': nan,\n",
       "    'running_time': '0.052 s'},\n",
       "   'capper': {'caps': {'income': 20000.0},\n",
       "    'transformed_column': ['income'],\n",
       "    'precomputed_caps': {'income': 20000.0},\n",
       "    'running_time': '0.049 s'},\n",
       "   'imputer': {'impute_strategy': 'median',\n",
       "    'columns_to_impute': ['bureau_score'],\n",
       "    'training_proportion_of_nulls': {'bureau_score': 0.10143635463100545},\n",
       "    'statistics': array([326.11326198]),\n",
       "    'running_time': '0.032 s'},\n",
       "   'lgbm_regression_learner': {'features': ['income',\n",
       "     'bureau_score',\n",
       "     'spend_desire',\n",
       "     'random_noise',\n",
       "     'monthly_spend',\n",
       "     'avg_last_2_months_spend'],\n",
       "    'target': 'target',\n",
       "    'prediction_column': 'prediction',\n",
       "    'package': 'lightgbm',\n",
       "    'package_version': '2.2.3',\n",
       "    'parameters': {'eta': 0.1,\n",
       "     'objective': 'regression',\n",
       "     'num_estimators': 200},\n",
       "    'feature_importance': {'income': 997,\n",
       "     'bureau_score': 965,\n",
       "     'spend_desire': 1028,\n",
       "     'random_noise': 907,\n",
       "     'monthly_spend': 895,\n",
       "     'avg_last_2_months_spend': 1208},\n",
       "    'training_samples': 10095,\n",
       "    'running_time': '7.943 s'},\n",
       "   'object': <lightgbm.basic.Booster at 0x1a2263c828>,\n",
       "   'prediction_ranger': {'prediction_min': 0.0,\n",
       "    'prediction_max': 100000.0,\n",
       "    'transformed_column': ['prediction']},\n",
       "   '__fkml__': {'pipeline': ['label_categorizer',\n",
       "     'capper',\n",
       "     'imputer',\n",
       "     'lgbm_regression_learner',\n",
       "     'prediction_ranger'],\n",
       "    'output_columns': ['id',\n",
       "     'month',\n",
       "     'income',\n",
       "     'created_at',\n",
       "     'phone_type',\n",
       "     'bureau_score',\n",
       "     'spend_desire',\n",
       "     'random_noise',\n",
       "     'monthly_spend',\n",
       "     'month_date',\n",
       "     'avg_last_2_months_spend',\n",
       "     'target',\n",
       "     'prediction'],\n",
       "    'features': ['id',\n",
       "     'month',\n",
       "     'income',\n",
       "     'created_at',\n",
       "     'phone_type',\n",
       "     'bureau_score',\n",
       "     'spend_desire',\n",
       "     'random_noise',\n",
       "     'monthly_spend',\n",
       "     'month_date',\n",
       "     'avg_last_2_months_spend',\n",
       "     'target'],\n",
       "    'learners': {'label_categorizer': {'fn': <function fklearn.training.transformation.label_categorizer.<locals>.p(new_df:pandas.core.frame.DataFrame) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'label_categorizer': {'transformed_column': ['phone_type'],\n",
       "        'replace_unseen': nan,\n",
       "        'running_time': '0.052 s'}}},\n",
       "     'capper': {'fn': <function fklearn.training.transformation.capper.<locals>.p(new_data_set:pandas.core.frame.DataFrame) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'capper': {'caps': {'income': 20000.0},\n",
       "        'transformed_column': ['income'],\n",
       "        'precomputed_caps': {'income': 20000.0},\n",
       "        'running_time': '0.049 s'}}},\n",
       "     'imputer': {'fn': <function fklearn.training.imputation.imputer.<locals>.p(new_data_set:pandas.core.frame.DataFrame) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'imputer': {'impute_strategy': 'median',\n",
       "        'columns_to_impute': ['bureau_score'],\n",
       "        'training_proportion_of_nulls': {'bureau_score': 0.10143635463100545},\n",
       "        'statistics': array([326.11326198]),\n",
       "        'running_time': '0.032 s'}}},\n",
       "     'lgbm_regression_learner': {'fn': <function fklearn.training.regression.lgbm_regression_learner.<locals>.p(new_df:pandas.core.frame.DataFrame, apply_shap:bool=False) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'lgbm_regression_learner': {'features': ['income',\n",
       "         'bureau_score',\n",
       "         'spend_desire',\n",
       "         'random_noise',\n",
       "         'monthly_spend',\n",
       "         'avg_last_2_months_spend'],\n",
       "        'target': 'target',\n",
       "        'prediction_column': 'prediction',\n",
       "        'package': 'lightgbm',\n",
       "        'package_version': '2.2.3',\n",
       "        'parameters': {'eta': 0.1,\n",
       "         'objective': 'regression',\n",
       "         'num_estimators': 200},\n",
       "        'feature_importance': {'income': 997,\n",
       "         'bureau_score': 965,\n",
       "         'spend_desire': 1028,\n",
       "         'random_noise': 907,\n",
       "         'monthly_spend': 895,\n",
       "         'avg_last_2_months_spend': 1208},\n",
       "        'training_samples': 10095,\n",
       "        'running_time': '7.943 s'},\n",
       "       'object': <lightgbm.basic.Booster at 0x1a2263c828>}},\n",
       "     'prediction_ranger': {'fn': <function fklearn.training.transformation.prediction_ranger.<locals>.p(new_df:pandas.core.frame.DataFrame) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'prediction_ranger': {'prediction_min': 0.0,\n",
       "        'prediction_max': 100000.0,\n",
       "        'transformed_column': ['prediction']}}}}}},\n",
       "  {'label_categorizer': {'transformed_column': ['phone_type'],\n",
       "    'replace_unseen': nan,\n",
       "    'running_time': '0.091 s'},\n",
       "   'capper': {'caps': {'income': 20000.0},\n",
       "    'transformed_column': ['income'],\n",
       "    'precomputed_caps': {'income': 20000.0},\n",
       "    'running_time': '0.025 s'},\n",
       "   'imputer': {'impute_strategy': 'median',\n",
       "    'columns_to_impute': ['bureau_score'],\n",
       "    'training_proportion_of_nulls': {'bureau_score': 0.09938140148058006},\n",
       "    'statistics': array([326.0731499]),\n",
       "    'running_time': '0.040 s'},\n",
       "   'lgbm_regression_learner': {'features': ['income',\n",
       "     'bureau_score',\n",
       "     'spend_desire',\n",
       "     'random_noise',\n",
       "     'monthly_spend',\n",
       "     'avg_last_2_months_spend'],\n",
       "    'target': 'target',\n",
       "    'prediction_column': 'prediction',\n",
       "    'package': 'lightgbm',\n",
       "    'package_version': '2.2.3',\n",
       "    'parameters': {'eta': 0.1,\n",
       "     'objective': 'regression',\n",
       "     'num_estimators': 200},\n",
       "    'feature_importance': {'income': 1025,\n",
       "     'bureau_score': 957,\n",
       "     'spend_desire': 1028,\n",
       "     'random_noise': 912,\n",
       "     'monthly_spend': 917,\n",
       "     'avg_last_2_months_spend': 1161},\n",
       "    'training_samples': 19722,\n",
       "    'running_time': '8.104 s'},\n",
       "   'object': <lightgbm.basic.Booster at 0x1a22647048>,\n",
       "   'prediction_ranger': {'prediction_min': 0.0,\n",
       "    'prediction_max': 100000.0,\n",
       "    'transformed_column': ['prediction']},\n",
       "   '__fkml__': {'pipeline': ['label_categorizer',\n",
       "     'capper',\n",
       "     'imputer',\n",
       "     'lgbm_regression_learner',\n",
       "     'prediction_ranger'],\n",
       "    'output_columns': ['id',\n",
       "     'month',\n",
       "     'income',\n",
       "     'created_at',\n",
       "     'phone_type',\n",
       "     'bureau_score',\n",
       "     'spend_desire',\n",
       "     'random_noise',\n",
       "     'monthly_spend',\n",
       "     'month_date',\n",
       "     'avg_last_2_months_spend',\n",
       "     'target',\n",
       "     'prediction'],\n",
       "    'features': ['id',\n",
       "     'month',\n",
       "     'income',\n",
       "     'created_at',\n",
       "     'phone_type',\n",
       "     'bureau_score',\n",
       "     'spend_desire',\n",
       "     'random_noise',\n",
       "     'monthly_spend',\n",
       "     'month_date',\n",
       "     'avg_last_2_months_spend',\n",
       "     'target'],\n",
       "    'learners': {'label_categorizer': {'fn': <function fklearn.training.transformation.label_categorizer.<locals>.p(new_df:pandas.core.frame.DataFrame) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'label_categorizer': {'transformed_column': ['phone_type'],\n",
       "        'replace_unseen': nan,\n",
       "        'running_time': '0.091 s'}}},\n",
       "     'capper': {'fn': <function fklearn.training.transformation.capper.<locals>.p(new_data_set:pandas.core.frame.DataFrame) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'capper': {'caps': {'income': 20000.0},\n",
       "        'transformed_column': ['income'],\n",
       "        'precomputed_caps': {'income': 20000.0},\n",
       "        'running_time': '0.025 s'}}},\n",
       "     'imputer': {'fn': <function fklearn.training.imputation.imputer.<locals>.p(new_data_set:pandas.core.frame.DataFrame) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'imputer': {'impute_strategy': 'median',\n",
       "        'columns_to_impute': ['bureau_score'],\n",
       "        'training_proportion_of_nulls': {'bureau_score': 0.09938140148058006},\n",
       "        'statistics': array([326.0731499]),\n",
       "        'running_time': '0.040 s'}}},\n",
       "     'lgbm_regression_learner': {'fn': <function fklearn.training.regression.lgbm_regression_learner.<locals>.p(new_df:pandas.core.frame.DataFrame, apply_shap:bool=False) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'lgbm_regression_learner': {'features': ['income',\n",
       "         'bureau_score',\n",
       "         'spend_desire',\n",
       "         'random_noise',\n",
       "         'monthly_spend',\n",
       "         'avg_last_2_months_spend'],\n",
       "        'target': 'target',\n",
       "        'prediction_column': 'prediction',\n",
       "        'package': 'lightgbm',\n",
       "        'package_version': '2.2.3',\n",
       "        'parameters': {'eta': 0.1,\n",
       "         'objective': 'regression',\n",
       "         'num_estimators': 200},\n",
       "        'feature_importance': {'income': 1025,\n",
       "         'bureau_score': 957,\n",
       "         'spend_desire': 1028,\n",
       "         'random_noise': 912,\n",
       "         'monthly_spend': 917,\n",
       "         'avg_last_2_months_spend': 1161},\n",
       "        'training_samples': 19722,\n",
       "        'running_time': '8.104 s'},\n",
       "       'object': <lightgbm.basic.Booster at 0x1a22647048>}},\n",
       "     'prediction_ranger': {'fn': <function fklearn.training.transformation.prediction_ranger.<locals>.p(new_df:pandas.core.frame.DataFrame) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'prediction_ranger': {'prediction_min': 0.0,\n",
       "        'prediction_max': 100000.0,\n",
       "        'transformed_column': ['prediction']}}}}}},\n",
       "  {'label_categorizer': {'transformed_column': ['phone_type'],\n",
       "    'replace_unseen': nan,\n",
       "    'running_time': '0.077 s'},\n",
       "   'capper': {'caps': {'income': 20000.0},\n",
       "    'transformed_column': ['income'],\n",
       "    'precomputed_caps': {'income': 20000.0},\n",
       "    'running_time': '0.023 s'},\n",
       "   'imputer': {'impute_strategy': 'median',\n",
       "    'columns_to_impute': ['bureau_score'],\n",
       "    'training_proportion_of_nulls': {'bureau_score': 0.09925883896592078},\n",
       "    'statistics': array([324.7631039]),\n",
       "    'running_time': '0.060 s'},\n",
       "   'lgbm_regression_learner': {'features': ['income',\n",
       "     'bureau_score',\n",
       "     'spend_desire',\n",
       "     'random_noise',\n",
       "     'monthly_spend',\n",
       "     'avg_last_2_months_spend'],\n",
       "    'target': 'target',\n",
       "    'prediction_column': 'prediction',\n",
       "    'package': 'lightgbm',\n",
       "    'package_version': '2.2.3',\n",
       "    'parameters': {'eta': 0.1,\n",
       "     'objective': 'regression',\n",
       "     'num_estimators': 200},\n",
       "    'feature_importance': {'income': 1031,\n",
       "     'bureau_score': 999,\n",
       "     'spend_desire': 1074,\n",
       "     'random_noise': 844,\n",
       "     'monthly_spend': 903,\n",
       "     'avg_last_2_months_spend': 1149},\n",
       "    'training_samples': 28199,\n",
       "    'running_time': '8.213 s'},\n",
       "   'object': <lightgbm.basic.Booster at 0x1a226472b0>,\n",
       "   'prediction_ranger': {'prediction_min': 0.0,\n",
       "    'prediction_max': 100000.0,\n",
       "    'transformed_column': ['prediction']},\n",
       "   '__fkml__': {'pipeline': ['label_categorizer',\n",
       "     'capper',\n",
       "     'imputer',\n",
       "     'lgbm_regression_learner',\n",
       "     'prediction_ranger'],\n",
       "    'output_columns': ['id',\n",
       "     'month',\n",
       "     'income',\n",
       "     'created_at',\n",
       "     'phone_type',\n",
       "     'bureau_score',\n",
       "     'spend_desire',\n",
       "     'random_noise',\n",
       "     'monthly_spend',\n",
       "     'month_date',\n",
       "     'avg_last_2_months_spend',\n",
       "     'target',\n",
       "     'prediction'],\n",
       "    'features': ['id',\n",
       "     'month',\n",
       "     'income',\n",
       "     'created_at',\n",
       "     'phone_type',\n",
       "     'bureau_score',\n",
       "     'spend_desire',\n",
       "     'random_noise',\n",
       "     'monthly_spend',\n",
       "     'month_date',\n",
       "     'avg_last_2_months_spend',\n",
       "     'target'],\n",
       "    'learners': {'label_categorizer': {'fn': <function fklearn.training.transformation.label_categorizer.<locals>.p(new_df:pandas.core.frame.DataFrame) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'label_categorizer': {'transformed_column': ['phone_type'],\n",
       "        'replace_unseen': nan,\n",
       "        'running_time': '0.077 s'}}},\n",
       "     'capper': {'fn': <function fklearn.training.transformation.capper.<locals>.p(new_data_set:pandas.core.frame.DataFrame) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'capper': {'caps': {'income': 20000.0},\n",
       "        'transformed_column': ['income'],\n",
       "        'precomputed_caps': {'income': 20000.0},\n",
       "        'running_time': '0.023 s'}}},\n",
       "     'imputer': {'fn': <function fklearn.training.imputation.imputer.<locals>.p(new_data_set:pandas.core.frame.DataFrame) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'imputer': {'impute_strategy': 'median',\n",
       "        'columns_to_impute': ['bureau_score'],\n",
       "        'training_proportion_of_nulls': {'bureau_score': 0.09925883896592078},\n",
       "        'statistics': array([324.7631039]),\n",
       "        'running_time': '0.060 s'}}},\n",
       "     'lgbm_regression_learner': {'fn': <function fklearn.training.regression.lgbm_regression_learner.<locals>.p(new_df:pandas.core.frame.DataFrame, apply_shap:bool=False) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'lgbm_regression_learner': {'features': ['income',\n",
       "         'bureau_score',\n",
       "         'spend_desire',\n",
       "         'random_noise',\n",
       "         'monthly_spend',\n",
       "         'avg_last_2_months_spend'],\n",
       "        'target': 'target',\n",
       "        'prediction_column': 'prediction',\n",
       "        'package': 'lightgbm',\n",
       "        'package_version': '2.2.3',\n",
       "        'parameters': {'eta': 0.1,\n",
       "         'objective': 'regression',\n",
       "         'num_estimators': 200},\n",
       "        'feature_importance': {'income': 1031,\n",
       "         'bureau_score': 999,\n",
       "         'spend_desire': 1074,\n",
       "         'random_noise': 844,\n",
       "         'monthly_spend': 903,\n",
       "         'avg_last_2_months_spend': 1149},\n",
       "        'training_samples': 28199,\n",
       "        'running_time': '8.213 s'},\n",
       "       'object': <lightgbm.basic.Booster at 0x1a226472b0>}},\n",
       "     'prediction_ranger': {'fn': <function fklearn.training.transformation.prediction_ranger.<locals>.p(new_df:pandas.core.frame.DataFrame) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'prediction_ranger': {'prediction_min': 0.0,\n",
       "        'prediction_max': 100000.0,\n",
       "        'transformed_column': ['prediction']}}}}}},\n",
       "  {'label_categorizer': {'transformed_column': ['phone_type'],\n",
       "    'replace_unseen': nan,\n",
       "    'running_time': '0.102 s'},\n",
       "   'capper': {'caps': {'income': 20000.0},\n",
       "    'transformed_column': ['income'],\n",
       "    'precomputed_caps': {'income': 20000.0},\n",
       "    'running_time': '0.026 s'},\n",
       "   'imputer': {'impute_strategy': 'median',\n",
       "    'columns_to_impute': ['bureau_score'],\n",
       "    'training_proportion_of_nulls': {'bureau_score': 0.09956821366274914},\n",
       "    'statistics': array([323.60346949]),\n",
       "    'running_time': '0.052 s'},\n",
       "   'lgbm_regression_learner': {'features': ['income',\n",
       "     'bureau_score',\n",
       "     'spend_desire',\n",
       "     'random_noise',\n",
       "     'monthly_spend',\n",
       "     'avg_last_2_months_spend'],\n",
       "    'target': 'target',\n",
       "    'prediction_column': 'prediction',\n",
       "    'package': 'lightgbm',\n",
       "    'package_version': '2.2.3',\n",
       "    'parameters': {'eta': 0.1,\n",
       "     'objective': 'regression',\n",
       "     'num_estimators': 200},\n",
       "    'feature_importance': {'income': 1031,\n",
       "     'bureau_score': 984,\n",
       "     'spend_desire': 1092,\n",
       "     'random_noise': 831,\n",
       "     'monthly_spend': 910,\n",
       "     'avg_last_2_months_spend': 1152},\n",
       "    'training_samples': 34971,\n",
       "    'running_time': '8.448 s'},\n",
       "   'object': <lightgbm.basic.Booster at 0x1a226473c8>,\n",
       "   'prediction_ranger': {'prediction_min': 0.0,\n",
       "    'prediction_max': 100000.0,\n",
       "    'transformed_column': ['prediction']},\n",
       "   '__fkml__': {'pipeline': ['label_categorizer',\n",
       "     'capper',\n",
       "     'imputer',\n",
       "     'lgbm_regression_learner',\n",
       "     'prediction_ranger'],\n",
       "    'output_columns': ['id',\n",
       "     'month',\n",
       "     'income',\n",
       "     'created_at',\n",
       "     'phone_type',\n",
       "     'bureau_score',\n",
       "     'spend_desire',\n",
       "     'random_noise',\n",
       "     'monthly_spend',\n",
       "     'month_date',\n",
       "     'avg_last_2_months_spend',\n",
       "     'target',\n",
       "     'prediction'],\n",
       "    'features': ['id',\n",
       "     'month',\n",
       "     'income',\n",
       "     'created_at',\n",
       "     'phone_type',\n",
       "     'bureau_score',\n",
       "     'spend_desire',\n",
       "     'random_noise',\n",
       "     'monthly_spend',\n",
       "     'month_date',\n",
       "     'avg_last_2_months_spend',\n",
       "     'target'],\n",
       "    'learners': {'label_categorizer': {'fn': <function fklearn.training.transformation.label_categorizer.<locals>.p(new_df:pandas.core.frame.DataFrame) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'label_categorizer': {'transformed_column': ['phone_type'],\n",
       "        'replace_unseen': nan,\n",
       "        'running_time': '0.102 s'}}},\n",
       "     'capper': {'fn': <function fklearn.training.transformation.capper.<locals>.p(new_data_set:pandas.core.frame.DataFrame) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'capper': {'caps': {'income': 20000.0},\n",
       "        'transformed_column': ['income'],\n",
       "        'precomputed_caps': {'income': 20000.0},\n",
       "        'running_time': '0.026 s'}}},\n",
       "     'imputer': {'fn': <function fklearn.training.imputation.imputer.<locals>.p(new_data_set:pandas.core.frame.DataFrame) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'imputer': {'impute_strategy': 'median',\n",
       "        'columns_to_impute': ['bureau_score'],\n",
       "        'training_proportion_of_nulls': {'bureau_score': 0.09956821366274914},\n",
       "        'statistics': array([323.60346949]),\n",
       "        'running_time': '0.052 s'}}},\n",
       "     'lgbm_regression_learner': {'fn': <function fklearn.training.regression.lgbm_regression_learner.<locals>.p(new_df:pandas.core.frame.DataFrame, apply_shap:bool=False) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'lgbm_regression_learner': {'features': ['income',\n",
       "         'bureau_score',\n",
       "         'spend_desire',\n",
       "         'random_noise',\n",
       "         'monthly_spend',\n",
       "         'avg_last_2_months_spend'],\n",
       "        'target': 'target',\n",
       "        'prediction_column': 'prediction',\n",
       "        'package': 'lightgbm',\n",
       "        'package_version': '2.2.3',\n",
       "        'parameters': {'eta': 0.1,\n",
       "         'objective': 'regression',\n",
       "         'num_estimators': 200},\n",
       "        'feature_importance': {'income': 1031,\n",
       "         'bureau_score': 984,\n",
       "         'spend_desire': 1092,\n",
       "         'random_noise': 831,\n",
       "         'monthly_spend': 910,\n",
       "         'avg_last_2_months_spend': 1152},\n",
       "        'training_samples': 34971,\n",
       "        'running_time': '8.448 s'},\n",
       "       'object': <lightgbm.basic.Booster at 0x1a226473c8>}},\n",
       "     'prediction_ranger': {'fn': <function fklearn.training.transformation.prediction_ranger.<locals>.p(new_df:pandas.core.frame.DataFrame) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'prediction_ranger': {'prediction_min': 0.0,\n",
       "        'prediction_max': 100000.0,\n",
       "        'transformed_column': ['prediction']}}}}}},\n",
       "  {'label_categorizer': {'transformed_column': ['phone_type'],\n",
       "    'replace_unseen': nan,\n",
       "    'running_time': '0.074 s'},\n",
       "   'capper': {'caps': {'income': 20000.0},\n",
       "    'transformed_column': ['income'],\n",
       "    'precomputed_caps': {'income': 20000.0},\n",
       "    'running_time': '0.023 s'},\n",
       "   'imputer': {'impute_strategy': 'median',\n",
       "    'columns_to_impute': ['bureau_score'],\n",
       "    'training_proportion_of_nulls': {'bureau_score': 0.10037035049111695},\n",
       "    'statistics': array([322.60691645]),\n",
       "    'running_time': '0.063 s'},\n",
       "   'lgbm_regression_learner': {'features': ['income',\n",
       "     'bureau_score',\n",
       "     'spend_desire',\n",
       "     'random_noise',\n",
       "     'monthly_spend',\n",
       "     'avg_last_2_months_spend'],\n",
       "    'target': 'target',\n",
       "    'prediction_column': 'prediction',\n",
       "    'package': 'lightgbm',\n",
       "    'package_version': '2.2.3',\n",
       "    'parameters': {'eta': 0.1,\n",
       "     'objective': 'regression',\n",
       "     'num_estimators': 200},\n",
       "    'feature_importance': {'income': 1032,\n",
       "     'bureau_score': 973,\n",
       "     'spend_desire': 1118,\n",
       "     'random_noise': 814,\n",
       "     'monthly_spend': 947,\n",
       "     'avg_last_2_months_spend': 1116},\n",
       "    'training_samples': 37262,\n",
       "    'running_time': '8.429 s'},\n",
       "   'object': <lightgbm.basic.Booster at 0x1a2263cc18>,\n",
       "   'prediction_ranger': {'prediction_min': 0.0,\n",
       "    'prediction_max': 100000.0,\n",
       "    'transformed_column': ['prediction']},\n",
       "   '__fkml__': {'pipeline': ['label_categorizer',\n",
       "     'capper',\n",
       "     'imputer',\n",
       "     'lgbm_regression_learner',\n",
       "     'prediction_ranger'],\n",
       "    'output_columns': ['id',\n",
       "     'month',\n",
       "     'income',\n",
       "     'created_at',\n",
       "     'phone_type',\n",
       "     'bureau_score',\n",
       "     'spend_desire',\n",
       "     'random_noise',\n",
       "     'monthly_spend',\n",
       "     'month_date',\n",
       "     'avg_last_2_months_spend',\n",
       "     'target',\n",
       "     'prediction'],\n",
       "    'features': ['id',\n",
       "     'month',\n",
       "     'income',\n",
       "     'created_at',\n",
       "     'phone_type',\n",
       "     'bureau_score',\n",
       "     'spend_desire',\n",
       "     'random_noise',\n",
       "     'monthly_spend',\n",
       "     'month_date',\n",
       "     'avg_last_2_months_spend',\n",
       "     'target'],\n",
       "    'learners': {'label_categorizer': {'fn': <function fklearn.training.transformation.label_categorizer.<locals>.p(new_df:pandas.core.frame.DataFrame) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'label_categorizer': {'transformed_column': ['phone_type'],\n",
       "        'replace_unseen': nan,\n",
       "        'running_time': '0.074 s'}}},\n",
       "     'capper': {'fn': <function fklearn.training.transformation.capper.<locals>.p(new_data_set:pandas.core.frame.DataFrame) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'capper': {'caps': {'income': 20000.0},\n",
       "        'transformed_column': ['income'],\n",
       "        'precomputed_caps': {'income': 20000.0},\n",
       "        'running_time': '0.023 s'}}},\n",
       "     'imputer': {'fn': <function fklearn.training.imputation.imputer.<locals>.p(new_data_set:pandas.core.frame.DataFrame) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'imputer': {'impute_strategy': 'median',\n",
       "        'columns_to_impute': ['bureau_score'],\n",
       "        'training_proportion_of_nulls': {'bureau_score': 0.10037035049111695},\n",
       "        'statistics': array([322.60691645]),\n",
       "        'running_time': '0.063 s'}}},\n",
       "     'lgbm_regression_learner': {'fn': <function fklearn.training.regression.lgbm_regression_learner.<locals>.p(new_df:pandas.core.frame.DataFrame, apply_shap:bool=False) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'lgbm_regression_learner': {'features': ['income',\n",
       "         'bureau_score',\n",
       "         'spend_desire',\n",
       "         'random_noise',\n",
       "         'monthly_spend',\n",
       "         'avg_last_2_months_spend'],\n",
       "        'target': 'target',\n",
       "        'prediction_column': 'prediction',\n",
       "        'package': 'lightgbm',\n",
       "        'package_version': '2.2.3',\n",
       "        'parameters': {'eta': 0.1,\n",
       "         'objective': 'regression',\n",
       "         'num_estimators': 200},\n",
       "        'feature_importance': {'income': 1032,\n",
       "         'bureau_score': 973,\n",
       "         'spend_desire': 1118,\n",
       "         'random_noise': 814,\n",
       "         'monthly_spend': 947,\n",
       "         'avg_last_2_months_spend': 1116},\n",
       "        'training_samples': 37262,\n",
       "        'running_time': '8.429 s'},\n",
       "       'object': <lightgbm.basic.Booster at 0x1a2263cc18>}},\n",
       "     'prediction_ranger': {'fn': <function fklearn.training.transformation.prediction_ranger.<locals>.p(new_df:pandas.core.frame.DataFrame) -> pandas.core.frame.DataFrame>,\n",
       "      'log': {'prediction_ranger': {'prediction_min': 0.0,\n",
       "        'prediction_max': 100000.0,\n",
       "        'transformed_column': ['prediction']}}}}}}],\n",
       " 'validator_log': [{'fold_num': 0,\n",
       "   'eval_results': [{'r2_evaluator__target': 0.5086378476205488,\n",
       "     'spearman_evaluator__target': 0.7386593691328156}],\n",
       "   'split_log': {'train_start': Timestamp('2017-02-01 00:00:00'),\n",
       "    'train_end': Timestamp('2018-03-11 00:00:00'),\n",
       "    'train_size': 5169,\n",
       "    'test_start': Timestamp('2018-04-11 00:00:00'),\n",
       "    'test_end': Timestamp('2018-06-12 00:00:00'),\n",
       "    'test_size': 16944,\n",
       "    'percentage': 0.1}},\n",
       "  {'fold_num': 1,\n",
       "   'eval_results': [{'r2_evaluator__target': 0.528636837654197,\n",
       "     'spearman_evaluator__target': 0.7559395783320386}],\n",
       "   'split_log': {'train_start': Timestamp('2017-02-01 00:00:00'),\n",
       "    'train_end': Timestamp('2018-03-11 00:00:00'),\n",
       "    'train_size': 10095,\n",
       "    'test_start': Timestamp('2018-04-11 00:00:00'),\n",
       "    'test_end': Timestamp('2018-06-12 00:00:00'),\n",
       "    'test_size': 16944,\n",
       "    'percentage': 0.2}},\n",
       "  {'fold_num': 2,\n",
       "   'eval_results': [{'r2_evaluator__target': 0.5526850769787538,\n",
       "     'spearman_evaluator__target': 0.7708684906608395}],\n",
       "   'split_log': {'train_start': Timestamp('2017-02-01 00:00:00'),\n",
       "    'train_end': Timestamp('2018-03-11 00:00:00'),\n",
       "    'train_size': 19722,\n",
       "    'test_start': Timestamp('2018-04-11 00:00:00'),\n",
       "    'test_end': Timestamp('2018-06-12 00:00:00'),\n",
       "    'test_size': 16944,\n",
       "    'percentage': 0.4}},\n",
       "  {'fold_num': 3,\n",
       "   'eval_results': [{'r2_evaluator__target': 0.5590326161307089,\n",
       "     'spearman_evaluator__target': 0.7746842469162849}],\n",
       "   'split_log': {'train_start': Timestamp('2017-02-01 00:00:00'),\n",
       "    'train_end': Timestamp('2018-03-11 00:00:00'),\n",
       "    'train_size': 28199,\n",
       "    'test_start': Timestamp('2018-04-11 00:00:00'),\n",
       "    'test_end': Timestamp('2018-06-12 00:00:00'),\n",
       "    'test_size': 16944,\n",
       "    'percentage': 0.6}},\n",
       "  {'fold_num': 4,\n",
       "   'eval_results': [{'r2_evaluator__target': 0.5610652544945294,\n",
       "     'spearman_evaluator__target': 0.7753973663995514}],\n",
       "   'split_log': {'train_start': Timestamp('2017-02-01 00:00:00'),\n",
       "    'train_end': Timestamp('2018-03-11 00:00:00'),\n",
       "    'train_size': 34971,\n",
       "    'test_start': Timestamp('2018-04-11 00:00:00'),\n",
       "    'test_end': Timestamp('2018-06-12 00:00:00'),\n",
       "    'test_size': 16944,\n",
       "    'percentage': 0.8}},\n",
       "  {'fold_num': 5,\n",
       "   'eval_results': [{'r2_evaluator__target': 0.5623244757695537,\n",
       "     'spearman_evaluator__target': 0.7760550988045266}],\n",
       "   'split_log': {'train_start': Timestamp('2017-02-01 00:00:00'),\n",
       "    'train_end': Timestamp('2018-03-11 00:00:00'),\n",
       "    'train_size': 37262,\n",
       "    'test_start': Timestamp('2018-04-11 00:00:00'),\n",
       "    'test_end': Timestamp('2018-06-12 00:00:00'),\n",
       "    'test_size': 16944,\n",
       "    'percentage': 1.0}}]}"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "spatial_learning_curve_logs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:50:12.804699Z",
     "start_time": "2019-09-26T20:50:12.745600Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "data = extract(spatial_learning_curve_logs['validator_log'], full_extractor)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:50:12.820352Z",
     "start_time": "2019-09-26T20:50:12.807168Z"
    },
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "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>r2_evaluator__target</th>\n",
       "      <th>spearman_evaluator__target</th>\n",
       "      <th>fold_num</th>\n",
       "      <th>train_start</th>\n",
       "      <th>train_end</th>\n",
       "      <th>train_size</th>\n",
       "      <th>test_start</th>\n",
       "      <th>test_end</th>\n",
       "      <th>test_size</th>\n",
       "      <th>percentage</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.508638</td>\n",
       "      <td>0.738659</td>\n",
       "      <td>0</td>\n",
       "      <td>2017-02-01</td>\n",
       "      <td>2018-03-11</td>\n",
       "      <td>5169</td>\n",
       "      <td>2018-04-11</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>16944</td>\n",
       "      <td>0.1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.528637</td>\n",
       "      <td>0.755940</td>\n",
       "      <td>1</td>\n",
       "      <td>2017-02-01</td>\n",
       "      <td>2018-03-11</td>\n",
       "      <td>10095</td>\n",
       "      <td>2018-04-11</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>16944</td>\n",
       "      <td>0.2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.552685</td>\n",
       "      <td>0.770868</td>\n",
       "      <td>2</td>\n",
       "      <td>2017-02-01</td>\n",
       "      <td>2018-03-11</td>\n",
       "      <td>19722</td>\n",
       "      <td>2018-04-11</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>16944</td>\n",
       "      <td>0.4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.559033</td>\n",
       "      <td>0.774684</td>\n",
       "      <td>3</td>\n",
       "      <td>2017-02-01</td>\n",
       "      <td>2018-03-11</td>\n",
       "      <td>28199</td>\n",
       "      <td>2018-04-11</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>16944</td>\n",
       "      <td>0.6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.561065</td>\n",
       "      <td>0.775397</td>\n",
       "      <td>4</td>\n",
       "      <td>2017-02-01</td>\n",
       "      <td>2018-03-11</td>\n",
       "      <td>34971</td>\n",
       "      <td>2018-04-11</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>16944</td>\n",
       "      <td>0.8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.562324</td>\n",
       "      <td>0.776055</td>\n",
       "      <td>5</td>\n",
       "      <td>2017-02-01</td>\n",
       "      <td>2018-03-11</td>\n",
       "      <td>37262</td>\n",
       "      <td>2018-04-11</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>16944</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   r2_evaluator__target  spearman_evaluator__target  fold_num train_start  \\\n",
       "0              0.508638                    0.738659         0  2017-02-01   \n",
       "0              0.528637                    0.755940         1  2017-02-01   \n",
       "0              0.552685                    0.770868         2  2017-02-01   \n",
       "0              0.559033                    0.774684         3  2017-02-01   \n",
       "0              0.561065                    0.775397         4  2017-02-01   \n",
       "0              0.562324                    0.776055         5  2017-02-01   \n",
       "\n",
       "   train_end  train_size test_start   test_end  test_size  percentage  \n",
       "0 2018-03-11        5169 2018-04-11 2018-06-12      16944         0.1  \n",
       "0 2018-03-11       10095 2018-04-11 2018-06-12      16944         0.2  \n",
       "0 2018-03-11       19722 2018-04-11 2018-06-12      16944         0.4  \n",
       "0 2018-03-11       28199 2018-04-11 2018-06-12      16944         0.6  \n",
       "0 2018-03-11       34971 2018-04-11 2018-06-12      16944         0.8  \n",
       "0 2018-03-11       37262 2018-04-11 2018-06-12      16944         1.0  "
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:50:13.030562Z",
     "start_time": "2019-09-26T20:50:12.822246Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(figsize=(10, 8))\n",
    "sns.pointplot(data=data, ax=ax, x=\"percentage\", y=\"r2_evaluator__target\")\n",
    "plt.title(\"Spatial learning curve (trained from %s to %s)\" % (data.train_start.dt.date.min(), data.train_end.dt.date.max()));"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-04-24T13:49:52.277446Z",
     "start_time": "2019-04-24T13:49:52.272394Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## Performance over time\n",
    "\n",
    "* <font size=\"4\"> That metric can be thought as split your dataset month by month and computing the desired metrics\n",
    "</font>\n",
    "\n",
    "\n",
    "* <font size=\"4\"> We can do that with ease using our split evaluators </font>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-26T20:50:16.123968Z",
     "start_time": "2019-09-26T20:50:16.114568Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "from fklearn.validation.evaluators import split_evaluator\n",
    "from fklearn.metrics.pd_extractors import split_evaluator_extractor\n",
    "out_of_space_holdout = pd.concat([scored_intime_outspace_hdout, scored_outime_outspace_hdout])\n",
    "\n",
    "monthly_eval_fn = split_evaluator(eval_fn=eval_fn,\n",
    "                                  split_col=\"month\",\n",
    "                                  split_values=list(range(0, 25)))\n",
    "\n",
    "monthly_extractor = split_evaluator_extractor(base_extractor=full_extractor, \n",
    "                                              split_col=\"month\", \n",
    "                                              split_values=list(range(0, 25)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-26T14:36:51.244424Z",
     "start_time": "2019-06-26T14:36:51.156741Z"
    },
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [],
   "source": [
    "out_of_space_logs = monthly_eval_fn(out_of_space_holdout)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-26T14:37:07.740922Z",
     "start_time": "2019-06-26T14:37:07.682994Z"
    },
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [],
   "source": [
    "monthly_performance = monthly_extractor(out_of_space_logs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-26T14:37:08.293833Z",
     "start_time": "2019-06-26T14:37:08.283570Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "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>r2_evaluator__target</th>\n",
       "      <th>spearman_evaluator__target</th>\n",
       "      <th>split_evaluator__month</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.421870</td>\n",
       "      <td>0.744979</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.491403</td>\n",
       "      <td>0.701534</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.478469</td>\n",
       "      <td>0.729370</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.534005</td>\n",
       "      <td>0.736146</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.538004</td>\n",
       "      <td>0.752896</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.528558</td>\n",
       "      <td>0.749593</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.599890</td>\n",
       "      <td>0.799352</td>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.603244</td>\n",
       "      <td>0.796775</td>\n",
       "      <td>8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.627977</td>\n",
       "      <td>0.808087</td>\n",
       "      <td>9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.598558</td>\n",
       "      <td>0.779260</td>\n",
       "      <td>10</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.604571</td>\n",
       "      <td>0.788846</td>\n",
       "      <td>11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.604477</td>\n",
       "      <td>0.793243</td>\n",
       "      <td>12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.575609</td>\n",
       "      <td>0.775225</td>\n",
       "      <td>13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.582989</td>\n",
       "      <td>0.789979</td>\n",
       "      <td>14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.578106</td>\n",
       "      <td>0.779769</td>\n",
       "      <td>15</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.576511</td>\n",
       "      <td>0.784794</td>\n",
       "      <td>16</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.611587</td>\n",
       "      <td>0.806708</td>\n",
       "      <td>17</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.616746</td>\n",
       "      <td>0.800946</td>\n",
       "      <td>18</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.581834</td>\n",
       "      <td>0.780355</td>\n",
       "      <td>19</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.576350</td>\n",
       "      <td>0.783299</td>\n",
       "      <td>20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>21</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>22</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>24</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   r2_evaluator__target  spearman_evaluator__target  split_evaluator__month\n",
       "0                   NaN                         NaN                       0\n",
       "0              0.421870                    0.744979                       1\n",
       "0              0.491403                    0.701534                       2\n",
       "0              0.478469                    0.729370                       3\n",
       "0              0.534005                    0.736146                       4\n",
       "0              0.538004                    0.752896                       5\n",
       "0              0.528558                    0.749593                       6\n",
       "0              0.599890                    0.799352                       7\n",
       "0              0.603244                    0.796775                       8\n",
       "0              0.627977                    0.808087                       9\n",
       "0              0.598558                    0.779260                      10\n",
       "0              0.604571                    0.788846                      11\n",
       "0              0.604477                    0.793243                      12\n",
       "0              0.575609                    0.775225                      13\n",
       "0              0.582989                    0.789979                      14\n",
       "0              0.578106                    0.779769                      15\n",
       "0              0.576511                    0.784794                      16\n",
       "0              0.611587                    0.806708                      17\n",
       "0              0.616746                    0.800946                      18\n",
       "0              0.581834                    0.780355                      19\n",
       "0              0.576350                    0.783299                      20\n",
       "0                   NaN                         NaN                      21\n",
       "0                   NaN                         NaN                      22\n",
       "0                   NaN                         NaN                      23\n",
       "0                   NaN                         NaN                      24"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "monthly_performance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-26T14:37:13.593862Z",
     "start_time": "2019-06-26T14:37:13.298620Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x1c240b51d0>"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(figsize=(10, 8))\n",
    "sns.pointplot(data=monthly_performance, ax=ax, x=\"split_evaluator__month\", y=\"r2_evaluator__target\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "##  Impact of More Data on Monthly Performance?\n",
    "\n",
    "* <font size=\"4\"> We want both our previous way of spliting our dataset and the evaluator we created </font>\n",
    "\n",
    "\n",
    "* <font size=\"4\"> We can use both to create a new spatial learning curve! </font>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-26T14:38:22.395792Z",
     "start_time": "2019-06-26T14:38:12.970873Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "monthly_spatial_learning_curve_logs = parallel_validator(train_set,\n",
    "                                                         split_fn,\n",
    "                                                         train_fn,\n",
    "                                                         monthly_eval_fn,\n",
    "                                                         n_jobs=8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-26T14:38:22.760243Z",
     "start_time": "2019-06-26T14:38:22.397990Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "monthly_data = extract(monthly_spatial_learning_curve_logs['validator_log'], monthly_extractor).loc[lambda df: df.r2_evaluator__target.notna()]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-26T14:38:22.780083Z",
     "start_time": "2019-06-26T14:38:22.762571Z"
    },
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "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>r2_evaluator__target</th>\n",
       "      <th>spearman_evaluator__target</th>\n",
       "      <th>split_evaluator__month</th>\n",
       "      <th>fold_num</th>\n",
       "      <th>train_start</th>\n",
       "      <th>train_end</th>\n",
       "      <th>train_size</th>\n",
       "      <th>test_start</th>\n",
       "      <th>test_end</th>\n",
       "      <th>test_size</th>\n",
       "      <th>percentage</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.540541</td>\n",
       "      <td>0.758164</td>\n",
       "      <td>15</td>\n",
       "      <td>0</td>\n",
       "      <td>2017-02-01</td>\n",
       "      <td>2018-03-11</td>\n",
       "      <td>5094</td>\n",
       "      <td>2018-04-11</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>16711</td>\n",
       "      <td>0.1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.514314</td>\n",
       "      <td>0.743269</td>\n",
       "      <td>16</td>\n",
       "      <td>0</td>\n",
       "      <td>2017-02-01</td>\n",
       "      <td>2018-03-11</td>\n",
       "      <td>5094</td>\n",
       "      <td>2018-04-11</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>16711</td>\n",
       "      <td>0.1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.507848</td>\n",
       "      <td>0.743218</td>\n",
       "      <td>17</td>\n",
       "      <td>0</td>\n",
       "      <td>2017-02-01</td>\n",
       "      <td>2018-03-11</td>\n",
       "      <td>5094</td>\n",
       "      <td>2018-04-11</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>16711</td>\n",
       "      <td>0.1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.566848</td>\n",
       "      <td>0.774746</td>\n",
       "      <td>15</td>\n",
       "      <td>1</td>\n",
       "      <td>2017-02-01</td>\n",
       "      <td>2018-03-11</td>\n",
       "      <td>9987</td>\n",
       "      <td>2018-04-11</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>16711</td>\n",
       "      <td>0.2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.543697</td>\n",
       "      <td>0.761993</td>\n",
       "      <td>16</td>\n",
       "      <td>1</td>\n",
       "      <td>2017-02-01</td>\n",
       "      <td>2018-03-11</td>\n",
       "      <td>9987</td>\n",
       "      <td>2018-04-11</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>16711</td>\n",
       "      <td>0.2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.539739</td>\n",
       "      <td>0.762033</td>\n",
       "      <td>17</td>\n",
       "      <td>1</td>\n",
       "      <td>2017-02-01</td>\n",
       "      <td>2018-03-11</td>\n",
       "      <td>9987</td>\n",
       "      <td>2018-04-11</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>16711</td>\n",
       "      <td>0.2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.586378</td>\n",
       "      <td>0.786420</td>\n",
       "      <td>15</td>\n",
       "      <td>2</td>\n",
       "      <td>2017-02-01</td>\n",
       "      <td>2018-03-11</td>\n",
       "      <td>19440</td>\n",
       "      <td>2018-04-11</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>16711</td>\n",
       "      <td>0.4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.565958</td>\n",
       "      <td>0.775002</td>\n",
       "      <td>16</td>\n",
       "      <td>2</td>\n",
       "      <td>2017-02-01</td>\n",
       "      <td>2018-03-11</td>\n",
       "      <td>19440</td>\n",
       "      <td>2018-04-11</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>16711</td>\n",
       "      <td>0.4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.562612</td>\n",
       "      <td>0.773501</td>\n",
       "      <td>17</td>\n",
       "      <td>2</td>\n",
       "      <td>2017-02-01</td>\n",
       "      <td>2018-03-11</td>\n",
       "      <td>19440</td>\n",
       "      <td>2018-04-11</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>16711</td>\n",
       "      <td>0.4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.593950</td>\n",
       "      <td>0.789789</td>\n",
       "      <td>15</td>\n",
       "      <td>3</td>\n",
       "      <td>2017-02-01</td>\n",
       "      <td>2018-03-11</td>\n",
       "      <td>28068</td>\n",
       "      <td>2018-04-11</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>16711</td>\n",
       "      <td>0.6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.572617</td>\n",
       "      <td>0.777124</td>\n",
       "      <td>16</td>\n",
       "      <td>3</td>\n",
       "      <td>2017-02-01</td>\n",
       "      <td>2018-03-11</td>\n",
       "      <td>28068</td>\n",
       "      <td>2018-04-11</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>16711</td>\n",
       "      <td>0.6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.567864</td>\n",
       "      <td>0.775599</td>\n",
       "      <td>17</td>\n",
       "      <td>3</td>\n",
       "      <td>2017-02-01</td>\n",
       "      <td>2018-03-11</td>\n",
       "      <td>28068</td>\n",
       "      <td>2018-04-11</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>16711</td>\n",
       "      <td>0.6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.597177</td>\n",
       "      <td>0.790955</td>\n",
       "      <td>15</td>\n",
       "      <td>4</td>\n",
       "      <td>2017-02-01</td>\n",
       "      <td>2018-03-11</td>\n",
       "      <td>34682</td>\n",
       "      <td>2018-04-11</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>16711</td>\n",
       "      <td>0.8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.577139</td>\n",
       "      <td>0.780944</td>\n",
       "      <td>16</td>\n",
       "      <td>4</td>\n",
       "      <td>2017-02-01</td>\n",
       "      <td>2018-03-11</td>\n",
       "      <td>34682</td>\n",
       "      <td>2018-04-11</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>16711</td>\n",
       "      <td>0.8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.570256</td>\n",
       "      <td>0.776765</td>\n",
       "      <td>17</td>\n",
       "      <td>4</td>\n",
       "      <td>2017-02-01</td>\n",
       "      <td>2018-03-11</td>\n",
       "      <td>34682</td>\n",
       "      <td>2018-04-11</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>16711</td>\n",
       "      <td>0.8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.597451</td>\n",
       "      <td>0.791407</td>\n",
       "      <td>15</td>\n",
       "      <td>5</td>\n",
       "      <td>2017-02-01</td>\n",
       "      <td>2018-03-11</td>\n",
       "      <td>36681</td>\n",
       "      <td>2018-04-11</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>16711</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.575909</td>\n",
       "      <td>0.779593</td>\n",
       "      <td>16</td>\n",
       "      <td>5</td>\n",
       "      <td>2017-02-01</td>\n",
       "      <td>2018-03-11</td>\n",
       "      <td>36681</td>\n",
       "      <td>2018-04-11</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>16711</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.570274</td>\n",
       "      <td>0.777338</td>\n",
       "      <td>17</td>\n",
       "      <td>5</td>\n",
       "      <td>2017-02-01</td>\n",
       "      <td>2018-03-11</td>\n",
       "      <td>36681</td>\n",
       "      <td>2018-04-11</td>\n",
       "      <td>2018-06-12</td>\n",
       "      <td>16711</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   r2_evaluator__target  spearman_evaluator__target  split_evaluator__month  \\\n",
       "0              0.540541                    0.758164                      15   \n",
       "0              0.514314                    0.743269                      16   \n",
       "0              0.507848                    0.743218                      17   \n",
       "0              0.566848                    0.774746                      15   \n",
       "0              0.543697                    0.761993                      16   \n",
       "0              0.539739                    0.762033                      17   \n",
       "0              0.586378                    0.786420                      15   \n",
       "0              0.565958                    0.775002                      16   \n",
       "0              0.562612                    0.773501                      17   \n",
       "0              0.593950                    0.789789                      15   \n",
       "0              0.572617                    0.777124                      16   \n",
       "0              0.567864                    0.775599                      17   \n",
       "0              0.597177                    0.790955                      15   \n",
       "0              0.577139                    0.780944                      16   \n",
       "0              0.570256                    0.776765                      17   \n",
       "0              0.597451                    0.791407                      15   \n",
       "0              0.575909                    0.779593                      16   \n",
       "0              0.570274                    0.777338                      17   \n",
       "\n",
       "   fold_num train_start  train_end  train_size test_start   test_end  \\\n",
       "0         0  2017-02-01 2018-03-11        5094 2018-04-11 2018-06-12   \n",
       "0         0  2017-02-01 2018-03-11        5094 2018-04-11 2018-06-12   \n",
       "0         0  2017-02-01 2018-03-11        5094 2018-04-11 2018-06-12   \n",
       "0         1  2017-02-01 2018-03-11        9987 2018-04-11 2018-06-12   \n",
       "0         1  2017-02-01 2018-03-11        9987 2018-04-11 2018-06-12   \n",
       "0         1  2017-02-01 2018-03-11        9987 2018-04-11 2018-06-12   \n",
       "0         2  2017-02-01 2018-03-11       19440 2018-04-11 2018-06-12   \n",
       "0         2  2017-02-01 2018-03-11       19440 2018-04-11 2018-06-12   \n",
       "0         2  2017-02-01 2018-03-11       19440 2018-04-11 2018-06-12   \n",
       "0         3  2017-02-01 2018-03-11       28068 2018-04-11 2018-06-12   \n",
       "0         3  2017-02-01 2018-03-11       28068 2018-04-11 2018-06-12   \n",
       "0         3  2017-02-01 2018-03-11       28068 2018-04-11 2018-06-12   \n",
       "0         4  2017-02-01 2018-03-11       34682 2018-04-11 2018-06-12   \n",
       "0         4  2017-02-01 2018-03-11       34682 2018-04-11 2018-06-12   \n",
       "0         4  2017-02-01 2018-03-11       34682 2018-04-11 2018-06-12   \n",
       "0         5  2017-02-01 2018-03-11       36681 2018-04-11 2018-06-12   \n",
       "0         5  2017-02-01 2018-03-11       36681 2018-04-11 2018-06-12   \n",
       "0         5  2017-02-01 2018-03-11       36681 2018-04-11 2018-06-12   \n",
       "\n",
       "   test_size  percentage  \n",
       "0      16711         0.1  \n",
       "0      16711         0.1  \n",
       "0      16711         0.1  \n",
       "0      16711         0.2  \n",
       "0      16711         0.2  \n",
       "0      16711         0.2  \n",
       "0      16711         0.4  \n",
       "0      16711         0.4  \n",
       "0      16711         0.4  \n",
       "0      16711         0.6  \n",
       "0      16711         0.6  \n",
       "0      16711         0.6  \n",
       "0      16711         0.8  \n",
       "0      16711         0.8  \n",
       "0      16711         0.8  \n",
       "0      16711         1.0  \n",
       "0      16711         1.0  \n",
       "0      16711         1.0  "
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "monthly_data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-26T14:38:45.764006Z",
     "start_time": "2019-06-26T14:38:45.506718Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(figsize=(10, 8))\n",
    "sns.pointplot(data=monthly_data, ax=ax, x=\"percentage\", y=\"r2_evaluator__target\", hue=\"split_evaluator__month\")\n",
    "plt.title(\"Spatial Learning Curve By Month\")\n",
    "plt.xticks(rotation=45);"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# What else does fklearn provide me?\n",
    "\n",
    "* <font size=\"4\"> Several other **learning curves** that can be used depending on what you want to evaluate </font>\n",
    "\n",
    "\n",
    "* <font size=\"4\"> Several other **algorithms** for models </font>\n",
    "\n",
    "\n",
    "* <font size=\"4\"> Other tools with similar interface for **feature selection, parameter tuning** </font>\n",
    "\n",
    "\n",
    "* <font size=\"4\"> All this methods are integrated with similar signatures in a way that is easy to reuse training, spliting and evaluation functions </font>\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Learn more:\n",
    "\n",
    "<font size=\"4\"> \n",
    "1. Medium blog post: https://medium.com/building-nubank/   \n",
    "    \n",
    "2. Documentation: https://fklearn.readthedocs.io/ </font>\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.8"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}