Data Science for Business - Chapter 3.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data Science for Business - Chapter 3\n",
    "> Introduction to Predictive Modeling: From Correlation to Supervised Segmentation\n",
    "\n",
    "### by Daenam Kim"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Outline\n",
    "1. Create a tree with IG based on the text book.\n",
    "1. Test"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 1. Create a tree with IG based on the text book."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import copy\n",
    "import math\n",
    "from random import randint\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "from IPython.display import display\n",
    "\n",
    "plt.style.use('seaborn-darkgrid')\n",
    "\n",
    "SHOW = True"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Define Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>age</th>\n",
       "      <th>balance</th>\n",
       "      <th>employed</th>\n",
       "      <th>residence</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>51</td>\n",
       "      <td>50120000</td>\n",
       "      <td>True</td>\n",
       "      <td>own</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>66</td>\n",
       "      <td>52062500</td>\n",
       "      <td>True</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>55</td>\n",
       "      <td>65350330</td>\n",
       "      <td>True</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>57</td>\n",
       "      <td>70200120</td>\n",
       "      <td>True</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>46</td>\n",
       "      <td>57009543</td>\n",
       "      <td>True</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>53</td>\n",
       "      <td>125605301</td>\n",
       "      <td>True</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>49</td>\n",
       "      <td>62103250</td>\n",
       "      <td>True</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>70</td>\n",
       "      <td>112510050</td>\n",
       "      <td>True</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>62</td>\n",
       "      <td>98445451</td>\n",
       "      <td>True</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>55</td>\n",
       "      <td>78540668</td>\n",
       "      <td>True</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>38</td>\n",
       "      <td>66580087</td>\n",
       "      <td>True</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>42</td>\n",
       "      <td>77655065</td>\n",
       "      <td>False</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>43</td>\n",
       "      <td>85465100</td>\n",
       "      <td>True</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>57</td>\n",
       "      <td>32151130</td>\n",
       "      <td>True</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>56</td>\n",
       "      <td>32215122</td>\n",
       "      <td>True</td>\n",
       "      <td>own</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>33</td>\n",
       "      <td>67560608</td>\n",
       "      <td>False</td>\n",
       "      <td>own</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>41</td>\n",
       "      <td>57951163</td>\n",
       "      <td>True</td>\n",
       "      <td>own</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>39</td>\n",
       "      <td>98750651</td>\n",
       "      <td>False</td>\n",
       "      <td>own</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>45</td>\n",
       "      <td>32135165</td>\n",
       "      <td>False</td>\n",
       "      <td>own</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>49</td>\n",
       "      <td>583151</td>\n",
       "      <td>False</td>\n",
       "      <td>own</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>37</td>\n",
       "      <td>99541</td>\n",
       "      <td>False</td>\n",
       "      <td>own</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>33</td>\n",
       "      <td>1121465</td>\n",
       "      <td>False</td>\n",
       "      <td>other</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>42</td>\n",
       "      <td>215521</td>\n",
       "      <td>False</td>\n",
       "      <td>other</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>41</td>\n",
       "      <td>1283544</td>\n",
       "      <td>False</td>\n",
       "      <td>other</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>35</td>\n",
       "      <td>2215985</td>\n",
       "      <td>False</td>\n",
       "      <td>other</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>28</td>\n",
       "      <td>5565587</td>\n",
       "      <td>False</td>\n",
       "      <td>other</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>32</td>\n",
       "      <td>992325</td>\n",
       "      <td>False</td>\n",
       "      <td>other</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>31</td>\n",
       "      <td>68848</td>\n",
       "      <td>False</td>\n",
       "      <td>other</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>29</td>\n",
       "      <td>984513</td>\n",
       "      <td>False</td>\n",
       "      <td>other</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>30</td>\n",
       "      <td>2565845</td>\n",
       "      <td>False</td>\n",
       "      <td>other</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    age    balance employed residence  label\n",
       "0    51   50120000     True       own   True\n",
       "1    66   52062500     True     other   True\n",
       "2    55   65350330     True     other   True\n",
       "3    57   70200120     True     other   True\n",
       "4    46   57009543     True     other   True\n",
       "5    53  125605301     True     other   True\n",
       "6    49   62103250     True     other   True\n",
       "7    70  112510050     True     other   True\n",
       "8    62   98445451     True     other   True\n",
       "9    55   78540668     True     other   True\n",
       "10   38   66580087     True     other   True\n",
       "11   42   77655065    False     other   True\n",
       "12   43   85465100     True     other   True\n",
       "13   57   32151130     True     other   True\n",
       "14   56   32215122     True       own  False\n",
       "15   33   67560608    False       own  False\n",
       "16   41   57951163     True       own  False\n",
       "17   39   98750651    False       own  False\n",
       "18   45   32135165    False       own  False\n",
       "19   49     583151    False       own  False\n",
       "20   37      99541    False       own  False\n",
       "21   33    1121465    False     other  False\n",
       "22   42     215521    False     other  False\n",
       "23   41    1283544    False     other  False\n",
       "24   35    2215985    False     other  False\n",
       "25   28    5565587    False     other  False\n",
       "26   32     992325    False     other  False\n",
       "27   31      68848    False     other  False\n",
       "28   29     984513    False     other  False\n",
       "29   30    2565845    False     other  False"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def get_learning_data():\n",
    "    \"\"\"\n",
    "    Return learning data.\n",
    "    \n",
    "    \"\"\"\n",
    "    inputs = [\n",
    "        {'age': 51, 'balance': 50120000, 'employed': 'True', 'residence': 'own', 'label': True},\n",
    "        {'age': 66, 'balance': 52062500, 'employed': 'True', 'residence': 'other', 'label': True},\n",
    "        {'age': 55, 'balance': 65350330, 'employed': 'True', 'residence': 'other', 'label': True},\n",
    "        {'age': 57, 'balance': 70200120, 'employed': 'True', 'residence': 'other', 'label': True},\n",
    "        {'age': 46, 'balance': 57009543, 'employed': 'True', 'residence': 'other', 'label': True},\n",
    "        {'age': 53, 'balance': 125605301, 'employed': 'True', 'residence': 'other', 'label': True},\n",
    "        {'age': 49, 'balance': 62103250, 'employed': 'True', 'residence': 'other', 'label': True},\n",
    "        {'age': 70, 'balance': 112510050, 'employed': 'True', 'residence': 'other', 'label': True},\n",
    "        {'age': 62, 'balance': 98445451, 'employed': 'True', 'residence': 'other', 'label': True},\n",
    "        {'age': 55, 'balance': 78540668, 'employed': 'True', 'residence': 'other', 'label': True},\n",
    "        {'age': 38, 'balance': 66580087, 'employed': 'True', 'residence': 'other', 'label': True},\n",
    "        {'age': 42, 'balance': 77655065, 'employed': 'False', 'residence': 'other', 'label': True},\n",
    "        {'age': 43, 'balance': 85465100, 'employed': 'True', 'residence': 'other', 'label': True},\n",
    "        {'age': 57, 'balance': 32151130, 'employed': 'True', 'residence': 'other', 'label': True},\n",
    "        {'age': 56, 'balance': 32215122, 'employed': 'True', 'residence': 'own', 'label': False},\n",
    "        {'age': 33, 'balance': 67560608, 'employed': 'False', 'residence': 'own', 'label': False},\n",
    "        {'age': 41, 'balance': 57951163, 'employed': 'True', 'residence': 'own', 'label': False},\n",
    "        {'age': 39, 'balance': 98750651, 'employed': 'False', 'residence': 'own', 'label': False},\n",
    "        {'age': 45, 'balance': 32135165, 'employed': 'False', 'residence': 'own', 'label': False},\n",
    "        {'age': 49, 'balance': 583151, 'employed': 'False', 'residence': 'own', 'label': False},\n",
    "        {'age': 37, 'balance': 99541, 'employed': 'False', 'residence': 'own', 'label': False},\n",
    "        {'age': 33, 'balance': 1121465, 'employed': 'False', 'residence': 'other', 'label': False},\n",
    "        {'age': 42, 'balance': 215521, 'employed': 'False', 'residence': 'other', 'label': False},\n",
    "        {'age': 41, 'balance': 1283544, 'employed': 'False', 'residence': 'other', 'label': False},\n",
    "        {'age': 35, 'balance': 2215985, 'employed': 'False', 'residence': 'other', 'label': False},\n",
    "        {'age': 28, 'balance': 5565587, 'employed': 'False', 'residence': 'other', 'label': False},\n",
    "        {'age': 32, 'balance': 992325, 'employed': 'False', 'residence': 'other', 'label': False},\n",
    "        {'age': 31, 'balance': 68848, 'employed': 'False', 'residence': 'other', 'label': False},\n",
    "        {'age': 29, 'balance': 984513, 'employed': 'False', 'residence': 'other', 'label': False},\n",
    "        {'age': 30, 'balance': 2565845, 'employed': 'False', 'residence': 'other', 'label': False}\n",
    "    ]\n",
    "    \n",
    "    return inputs\n",
    "\n",
    "\n",
    "# Labels. (aka Targets)\n",
    "LABELS = (True, False)\n",
    "\n",
    "# Attributes. (aka Variables)\n",
    "ATTRIBUTES = {\n",
    "    'age': ['int', 45],\n",
    "    'employed': ['str', 'True', 'False'],\n",
    "    'balance': ['int', 50000000],\n",
    "    'residence': ['str', 'own', 'other'],\n",
    "}\n",
    "\n",
    "INPUTS = pd.DataFrame(get_learning_data())\n",
    "INPUTS = INPUTS[['age', 'balance', 'employed', 'residence', 'label']]\n",
    "if SHOW:\n",
    "    display(INPUTS)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Attribute Selection for a Root Node based on the Information Gain"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def get_p(a, b):\n",
    "    len_a = len(a)\n",
    "    len_b = len(b)\n",
    "\n",
    "    if len_a == 0:\n",
    "        len_a = 0.00000001\n",
    "    \n",
    "    if len_b == 0:\n",
    "        len_b = 0.00000001\n",
    "        \n",
    "    return len_a / len_b\n",
    "\n",
    "def get_entropy_params(labels, data):\n",
    "    return (\n",
    "        get_p(data[data['label'] == labels[0]], data),\n",
    "        get_p(data[data['label'] == labels[1]], data)\n",
    "    )\n",
    "\n",
    "def get_entropy(labels, data):\n",
    "    ps = get_entropy_params(labels, data)\n",
    "    return -sum([p * math.log(p, 2) for p in ps])\n",
    "\n",
    "def get_information_gain(labels, attributes, inputs):\n",
    "    # Use labels only for a parent.\n",
    "    data = inputs[['label']]\n",
    "    entropy_parent = get_entropy(labels, data)\n",
    "\n",
    "    entropies = {'parent': entropy_parent}\n",
    "    igs = {}\n",
    "    for attr_name, attr_values in attributes.items():\n",
    "        data = inputs[[attr_name, 'label']]\n",
    "        mode = attr_values[0]\n",
    "        if mode == 'int':\n",
    "            attr_value = attr_values[1]\n",
    "            igs[attr_name] = entropy_parent - (\n",
    "                get_p(data[data[attr_name] < attr_value], data) * get_entropy(labels, data[data[attr_name] < attr_value]) +\n",
    "                get_p(data[data[attr_name] >= attr_value], data) * get_entropy(labels, data[data[attr_name] >= attr_value])\n",
    "            )\n",
    "            entropies[attr_name + '<' + str(attr_value)] = get_entropy(labels, data[data[attr_name] < attr_value])\n",
    "            entropies[attr_name + '>=' + str(attr_value)] = get_entropy(labels, data[data[attr_name] >= attr_value])\n",
    "        else:\n",
    "            children = []\n",
    "            for attr_value in attr_values[1:]:\n",
    "                children.append(get_p(data[data[attr_name] == attr_value], data) * get_entropy(labels, data[data[attr_name] == attr_value]))\n",
    "                entropies[attr_name + '_' + attr_value] = get_entropy(labels, data[data[attr_name] == attr_value])\n",
    "            \n",
    "            igs[attr_name] = entropy_parent - sum(children)\n",
    "\n",
    "    return igs, entropies"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Attribute</th>\n",
       "      <th>Entropy ( ↑ )</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>employed_False</td>\n",
       "      <td>0.353359</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>balance&lt;50000000</td>\n",
       "      <td>0.371232</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>residence_own</td>\n",
       "      <td>0.543564</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>employed_True</td>\n",
       "      <td>0.566510</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>age&lt;45</td>\n",
       "      <td>0.696212</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>balance&gt;=50000000</td>\n",
       "      <td>0.696212</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>age&gt;=45</td>\n",
       "      <td>0.749595</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>residence_other</td>\n",
       "      <td>0.976021</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>parent</td>\n",
       "      <td>0.996792</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           Attribute  Entropy ( ↑ )\n",
       "0     employed_False       0.353359\n",
       "1   balance<50000000       0.371232\n",
       "2      residence_own       0.543564\n",
       "3      employed_True       0.566510\n",
       "4             age<45       0.696212\n",
       "5  balance>=50000000       0.696212\n",
       "6            age>=45       0.749595\n",
       "7    residence_other       0.976021\n",
       "8             parent       0.996792"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "igs, entropies = get_information_gain(LABELS, ATTRIBUTES, INPUTS)\n",
    "entropies_sorted  = sorted(entropies.items(), key=lambda x: x[1])\n",
    "entropies_df = pd.DataFrame(entropies_sorted, columns=['Attribute', 'Entropy ( ↑ )'])\n",
    "if SHOW:\n",
    "    display(entropies_df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAacAAAFlCAYAAABPzycnAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ndDj/B/gL7PbPN6E0AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x111303d68>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.title('Entropy of Each Attribute', fontsize=18)\n",
    "plt.ylabel('Entropy', fontsize=14)\n",
    "plt.xlabel('Attribute', fontsize=14)\n",
    "plt.xticks(range(len(entropies)), entropies.keys(), rotation=-45, ha='left', fontsize=12)\n",
    "plt.bar(range(len(entropies)), entropies.values(), align='center')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Attribute</th>\n",
       "      <th>IG ( ↓ )</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>employed</td>\n",
       "      <td>0.536857</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>balance</td>\n",
       "      <td>0.452237</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>age</td>\n",
       "      <td>0.275667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>residence</td>\n",
       "      <td>0.136093</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Attribute  IG ( ↓ )\n",
       "0   employed  0.536857\n",
       "1    balance  0.452237\n",
       "2        age  0.275667\n",
       "3  residence  0.136093"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "igs_sorted  = sorted(igs.items(), key=lambda x: x[1], reverse=True)\n",
    "igs_df = pd.DataFrame(igs_sorted, columns=['Attribute', 'IG ( ↓ )'])\n",
    "if SHOW:\n",
    "    display(igs_df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYAAAAE9CAYAAAACk7UUAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x116354be0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.title('IG of Each Attribute for the Root Node ', fontsize=18)\n",
    "plt.ylabel('IG', fontsize=14)\n",
    "plt.xlabel('Attribute', fontsize=14)\n",
    "plt.xticks(range(len(igs)), igs.keys(), rotation=-45, ha='left', fontsize=12)\n",
    "plt.bar(range(len(igs)), igs.values(), align='center')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Devide-and-Conquer and Tree-Structured Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "class Tree():\n",
    "\n",
    "    def __init__(self, **kwargs):\n",
    "        self.left =  None\n",
    "        self.right = None\n",
    "        self.evaluation = 'Are you kidding me?'\n",
    "        self.update(**kwargs)\n",
    "\n",
    "    def add(self, **kwargs):\n",
    "        if kwargs['direction'] == 'left':\n",
    "            self.left = Tree(**kwargs)\n",
    "            return self.left\n",
    "        else:\n",
    "            self.right = Tree(**kwargs)\n",
    "            return self.right\n",
    "\n",
    "    def update(self, **kwargs):\n",
    "        self.id = kwargs.get('id', None)\n",
    "        self.key = kwargs.get('key', None)\n",
    "        self.values = kwargs.get('values', None)\n",
    "        self.mode =  kwargs.get('mode', None)\n",
    "        self.depth =  kwargs.get('depth', 0)\n",
    "        self.direction = kwargs.get('direction', None)\n",
    "        self.evaluation = kwargs.get('evaluation', None)\n",
    "        return self\n",
    "    \n",
    "    def evaluate(self, **kwargs):\n",
    "        if self.key is None:\n",
    "            return self.evaluation\n",
    "        \n",
    "        attr_value = kwargs[self.key]\n",
    "        if self.mode == 'int':\n",
    "            if attr_value < self.values[0]:\n",
    "                if self.left is not None:\n",
    "                    return self.left.evaluate(**kwargs)\n",
    "                else:\n",
    "                    return self.evaluation\n",
    "            else:\n",
    "                if self.right is not None:\n",
    "                    return self.right.evaluate(**kwargs)\n",
    "                else:\n",
    "                    return self.evaluation\n",
    "        else:\n",
    "            if attr_value == self.values[0]:\n",
    "                if self.left is not None:\n",
    "                    return self.left.evaluate(**kwargs)\n",
    "                else:\n",
    "                    return self.evaluation\n",
    "            else:\n",
    "                if self.right is not None:\n",
    "                    return self.right.evaluate(**kwargs)\n",
    "                else:\n",
    "                    return self.evaluation\n",
    "\n",
    "    def __str__(self):\n",
    "        if self.left != None:\n",
    "            self.left.__str__()\n",
    "\n",
    "        print({\n",
    "            'id': self.id,\n",
    "            'key': self.key,\n",
    "            'values': self.values,\n",
    "            'mode': self.mode,\n",
    "            'depth': self.depth,\n",
    "            'direction': self.direction,\n",
    "            'evaluation': self.evaluation\n",
    "        })\n",
    "\n",
    "        if self.right != None:\n",
    "            self.right.__str__()\n",
    "\n",
    "\n",
    "def is_write_off(labels, data):\n",
    "        no_write_off = len(data[data['label'] == labels[0]])\n",
    "        write_off = len(data[data['label'] == labels[1]])\n",
    "        \n",
    "        if no_write_off < 1 and write_off < 1:\n",
    "            return 'ERROR'\n",
    "        elif no_write_off > write_off:\n",
    "            return 'NO-WRITE-OFF'\n",
    "        else:\n",
    "            return 'WRITE-OFF'\n",
    "\n",
    "\n",
    "def get_terminal_node_data(node_id, depth, direction, labels, data):\n",
    "        return {\n",
    "            'id': node_id,\n",
    "            'key': None,\n",
    "            'values': None,\n",
    "            'mode': None,\n",
    "            'depth': depth,\n",
    "            'direction': direction,\n",
    "            'evaluation': is_write_off(labels, data)\n",
    "        }"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'id': 4, 'key': None, 'values': None, 'mode': None, 'depth': 2, 'direction': 'left', 'evaluation': 'WRITE-OFF'}\n",
      "{'id': 2, 'key': 'balance', 'values': [50000000], 'mode': 'int', 'depth': 1, 'direction': 'left', 'evaluation': None}\n",
      "{'id': 5, 'key': None, 'values': None, 'mode': None, 'depth': 2, 'direction': 'right', 'evaluation': 'NO-WRITE-OFF'}\n",
      "{'id': 1, 'key': 'employed', 'values': ['True', 'False'], 'mode': 'str', 'depth': 0, 'direction': None, 'evaluation': None}\n",
      "{'id': 3, 'key': None, 'values': None, 'mode': None, 'depth': 1, 'direction': 'right', 'evaluation': 'WRITE-OFF'}\n"
     ]
    }
   ],
   "source": [
    "def create_tree(tree, labels, attributes, inputs, depth=0, max_depth=10, node_id=None, direction=None):\n",
    "    if tree is None:\n",
    "        return\n",
    "\n",
    "    \"\"\"\n",
    "    Conditions to become a terminal node.\n",
    "    \n",
    "    1. At the entropy is smaller than \"entropy_limit\".\n",
    "    2. At no left or right data after dividing by the \"attribute\".\n",
    "    3. At the specific \"max_depth\".\n",
    "    \"\"\"\n",
    "    igs, entropies = get_information_gain(labels, attributes , inputs)\n",
    "    igs_sorted  = sorted(igs.items(), key=lambda x: x[1], reverse=True)\n",
    "    igs_max = igs_sorted[0]\n",
    "    attr_key = igs_max[0]\n",
    "    attr_values = attributes[attr_key]\n",
    "    mode = attr_values[0]\n",
    "    if mode == 'int':\n",
    "        v = attr_values[1]\n",
    "        inputs_left = inputs[inputs[attr_key] < v]\n",
    "        inputs_right = inputs[inputs[attr_key] >= v]\n",
    "    else:\n",
    "        inputs_left = inputs[inputs[attr_key] == attr_values[1]]\n",
    "        inputs_right = inputs[inputs[attr_key] == attr_values[2]]\n",
    "\n",
    "    inputs_left.reset_index(inplace=True)\n",
    "    inputs_right.reset_index(inplace=True)\n",
    "    del inputs_left['index']\n",
    "    del inputs_right['index']\n",
    "    if node_id is None:\n",
    "        node_id = 1\n",
    "\n",
    "    node_data = {\n",
    "        'id': node_id,\n",
    "        'key': attr_key,\n",
    "        'values': attr_values[1:],\n",
    "        'mode': mode,\n",
    "        'depth': depth,\n",
    "        'direction': direction,\n",
    "    }\n",
    "\n",
    "    entropy_limit = 0.1\n",
    "    \n",
    "    if (depth >= max_depth) or len(inputs_left) == 0 or len(inputs_right) == 0 or \\\n",
    "        get_entropy(labels, inputs_left) < entropy_limit or get_entropy(labels, inputs_right) < entropy_limit:\n",
    "\n",
    "        next_node = tree.add(**node_data)\n",
    "        next_node.update(**get_terminal_node_data(node_id, depth, direction, labels, inputs))\n",
    "        return\n",
    "    elif depth == 0:\n",
    "        next_node = tree.update(**node_data)\n",
    "    else:\n",
    "        next_node = tree.add(**node_data)\n",
    "\n",
    "    if depth == 0:\n",
    "        # Update attributes based on the biggest IG at the first time only.\n",
    "        next_attributes = {k: v for k, v in attributes.items() if k != attr_key}\n",
    "    else:\n",
    "        next_attributes = attributes\n",
    "    \n",
    "    create_tree(next_node, labels, next_attributes, inputs_left, depth + 1, max_depth, node_id * 2, 'left')\n",
    "    create_tree(next_node, labels, next_attributes, inputs_right, depth + 1, max_depth, (node_id * 2) + 1, 'right')\n",
    "        \n",
    "tree = Tree()\n",
    "# Remove \"residence\" attribute because its IG is lower than others.\n",
    "create_tree(tree, LABELS, {k: v for k, v in ATTRIBUTES.items() if k != 'residence'}, INPUTS)\n",
    "tree.__str__()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 2. Test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>age</th>\n",
       "      <th>balance</th>\n",
       "      <th>employed</th>\n",
       "      <th>residence</th>\n",
       "      <th>label</th>\n",
       "      <th>evaluation</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>51</td>\n",
       "      <td>50120000</td>\n",
       "      <td>True</td>\n",
       "      <td>own</td>\n",
       "      <td>True</td>\n",
       "      <td>NO-WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>66</td>\n",
       "      <td>52062500</td>\n",
       "      <td>True</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "      <td>NO-WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>55</td>\n",
       "      <td>65350330</td>\n",
       "      <td>True</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "      <td>NO-WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>57</td>\n",
       "      <td>70200120</td>\n",
       "      <td>True</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "      <td>NO-WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>46</td>\n",
       "      <td>57009543</td>\n",
       "      <td>True</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "      <td>NO-WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>53</td>\n",
       "      <td>125605301</td>\n",
       "      <td>True</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "      <td>NO-WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>49</td>\n",
       "      <td>62103250</td>\n",
       "      <td>True</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "      <td>NO-WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>70</td>\n",
       "      <td>112510050</td>\n",
       "      <td>True</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "      <td>NO-WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>62</td>\n",
       "      <td>98445451</td>\n",
       "      <td>True</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "      <td>NO-WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>55</td>\n",
       "      <td>78540668</td>\n",
       "      <td>True</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "      <td>NO-WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>38</td>\n",
       "      <td>66580087</td>\n",
       "      <td>True</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "      <td>NO-WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>42</td>\n",
       "      <td>77655065</td>\n",
       "      <td>False</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "      <td>WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>43</td>\n",
       "      <td>85465100</td>\n",
       "      <td>True</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "      <td>NO-WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>57</td>\n",
       "      <td>32151130</td>\n",
       "      <td>True</td>\n",
       "      <td>other</td>\n",
       "      <td>True</td>\n",
       "      <td>WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>56</td>\n",
       "      <td>32215122</td>\n",
       "      <td>True</td>\n",
       "      <td>own</td>\n",
       "      <td>False</td>\n",
       "      <td>WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>33</td>\n",
       "      <td>67560608</td>\n",
       "      <td>False</td>\n",
       "      <td>own</td>\n",
       "      <td>False</td>\n",
       "      <td>WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>41</td>\n",
       "      <td>57951163</td>\n",
       "      <td>True</td>\n",
       "      <td>own</td>\n",
       "      <td>False</td>\n",
       "      <td>NO-WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>39</td>\n",
       "      <td>98750651</td>\n",
       "      <td>False</td>\n",
       "      <td>own</td>\n",
       "      <td>False</td>\n",
       "      <td>WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>45</td>\n",
       "      <td>32135165</td>\n",
       "      <td>False</td>\n",
       "      <td>own</td>\n",
       "      <td>False</td>\n",
       "      <td>WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>49</td>\n",
       "      <td>583151</td>\n",
       "      <td>False</td>\n",
       "      <td>own</td>\n",
       "      <td>False</td>\n",
       "      <td>WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>37</td>\n",
       "      <td>99541</td>\n",
       "      <td>False</td>\n",
       "      <td>own</td>\n",
       "      <td>False</td>\n",
       "      <td>WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>33</td>\n",
       "      <td>1121465</td>\n",
       "      <td>False</td>\n",
       "      <td>other</td>\n",
       "      <td>False</td>\n",
       "      <td>WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>42</td>\n",
       "      <td>215521</td>\n",
       "      <td>False</td>\n",
       "      <td>other</td>\n",
       "      <td>False</td>\n",
       "      <td>WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>41</td>\n",
       "      <td>1283544</td>\n",
       "      <td>False</td>\n",
       "      <td>other</td>\n",
       "      <td>False</td>\n",
       "      <td>WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>35</td>\n",
       "      <td>2215985</td>\n",
       "      <td>False</td>\n",
       "      <td>other</td>\n",
       "      <td>False</td>\n",
       "      <td>WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>28</td>\n",
       "      <td>5565587</td>\n",
       "      <td>False</td>\n",
       "      <td>other</td>\n",
       "      <td>False</td>\n",
       "      <td>WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>32</td>\n",
       "      <td>992325</td>\n",
       "      <td>False</td>\n",
       "      <td>other</td>\n",
       "      <td>False</td>\n",
       "      <td>WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>31</td>\n",
       "      <td>68848</td>\n",
       "      <td>False</td>\n",
       "      <td>other</td>\n",
       "      <td>False</td>\n",
       "      <td>WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>29</td>\n",
       "      <td>984513</td>\n",
       "      <td>False</td>\n",
       "      <td>other</td>\n",
       "      <td>False</td>\n",
       "      <td>WRITE-OFF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>30</td>\n",
       "      <td>2565845</td>\n",
       "      <td>False</td>\n",
       "      <td>other</td>\n",
       "      <td>False</td>\n",
       "      <td>WRITE-OFF</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    age    balance employed residence  label    evaluation\n",
       "0    51   50120000     True       own   True  NO-WRITE-OFF\n",
       "1    66   52062500     True     other   True  NO-WRITE-OFF\n",
       "2    55   65350330     True     other   True  NO-WRITE-OFF\n",
       "3    57   70200120     True     other   True  NO-WRITE-OFF\n",
       "4    46   57009543     True     other   True  NO-WRITE-OFF\n",
       "5    53  125605301     True     other   True  NO-WRITE-OFF\n",
       "6    49   62103250     True     other   True  NO-WRITE-OFF\n",
       "7    70  112510050     True     other   True  NO-WRITE-OFF\n",
       "8    62   98445451     True     other   True  NO-WRITE-OFF\n",
       "9    55   78540668     True     other   True  NO-WRITE-OFF\n",
       "10   38   66580087     True     other   True  NO-WRITE-OFF\n",
       "11   42   77655065    False     other   True     WRITE-OFF\n",
       "12   43   85465100     True     other   True  NO-WRITE-OFF\n",
       "13   57   32151130     True     other   True     WRITE-OFF\n",
       "14   56   32215122     True       own  False     WRITE-OFF\n",
       "15   33   67560608    False       own  False     WRITE-OFF\n",
       "16   41   57951163     True       own  False  NO-WRITE-OFF\n",
       "17   39   98750651    False       own  False     WRITE-OFF\n",
       "18   45   32135165    False       own  False     WRITE-OFF\n",
       "19   49     583151    False       own  False     WRITE-OFF\n",
       "20   37      99541    False       own  False     WRITE-OFF\n",
       "21   33    1121465    False     other  False     WRITE-OFF\n",
       "22   42     215521    False     other  False     WRITE-OFF\n",
       "23   41    1283544    False     other  False     WRITE-OFF\n",
       "24   35    2215985    False     other  False     WRITE-OFF\n",
       "25   28    5565587    False     other  False     WRITE-OFF\n",
       "26   32     992325    False     other  False     WRITE-OFF\n",
       "27   31      68848    False     other  False     WRITE-OFF\n",
       "28   29     984513    False     other  False     WRITE-OFF\n",
       "29   30    2565845    False     other  False     WRITE-OFF"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Test for learning data.\n",
    "inputs = list(INPUTS.T.to_dict().values())\n",
    "for data in inputs:\n",
    "    data['evaluation'] = tree.evaluate(**data)\n",
    "\n",
    "evaluations = pd.DataFrame(inputs)\n",
    "evaluations = evaluations[['age', 'balance', 'employed', 'residence', 'label', 'evaluation']]\n",
    "display(evaluations)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy of LEARNING DATA is 0.9\n"
     ]
    }
   ],
   "source": [
    "no_write_off = evaluations[(evaluations['label'] == True) & (evaluations['evaluation'] == 'NO-WRITE-OFF')]\n",
    "write_off = evaluations[(evaluations['label'] == False) & (evaluations['evaluation'] == 'WRITE-OFF')]\n",
    "total = len(evaluations)\n",
    "matched = len(no_write_off) + len(write_off)\n",
    "print('Accuracy of LEARNING DATA is ' + str(matched / total))"
   ]
  }
 ],
 "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": "3.6.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}