problem1_1_notes.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Problem 1.1: Bank Marketing Data\n",
    "Aaron Bell"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## -1) Import core packages"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# for displaying plots in a notebook\n",
    "%matplotlib inline \n",
    "\n",
    "import numpy as np # simply array and mathematical operations\n",
    "import pandas as pd # R-like handling with dataframes, etc.\n",
    "import seaborn as sb # for making your statistical plots look pretty\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 0) Data loading\n",
    "pandas allows for easy reading-in of CSV and other files. \n",
    "We'll load the data into a pandas dataframe (analagous to R dataframes), then take a look at how it's stuctured.\n",
    "\n",
    "IMPORTANT: 'read_csv' sometimes gets confused if the file is delimited in a weird way. In this case, the columns are separated by semi-colon rather than comma, se we have to set 'sep = \";\"'. Otherwise all the data go to one column.\n",
    "\n",
    "We also know from the instructions (copied at the end of this notebook) that the target label is column 'y'. This corresponds to is whether customers said \"yes\" or \"no\" to a term deposit."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>age</th>\n",
       "      <th>job</th>\n",
       "      <th>marital</th>\n",
       "      <th>education</th>\n",
       "      <th>default</th>\n",
       "      <th>housing</th>\n",
       "      <th>loan</th>\n",
       "      <th>contact</th>\n",
       "      <th>month</th>\n",
       "      <th>day_of_week</th>\n",
       "      <th>duration</th>\n",
       "      <th>campaign</th>\n",
       "      <th>pdays</th>\n",
       "      <th>previous</th>\n",
       "      <th>poutcome</th>\n",
       "      <th>emp.var.rate</th>\n",
       "      <th>cons.price.idx</th>\n",
       "      <th>cons.conf.idx</th>\n",
       "      <th>euribor3m</th>\n",
       "      <th>nr.employed</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>56</td>\n",
       "      <td>housemaid</td>\n",
       "      <td>married</td>\n",
       "      <td>basic.4y</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>telephone</td>\n",
       "      <td>may</td>\n",
       "      <td>mon</td>\n",
       "      <td>261</td>\n",
       "      <td>1</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.1</td>\n",
       "      <td>93.994</td>\n",
       "      <td>-36.4</td>\n",
       "      <td>4.857</td>\n",
       "      <td>5191.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>57</td>\n",
       "      <td>services</td>\n",
       "      <td>married</td>\n",
       "      <td>high.school</td>\n",
       "      <td>unknown</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>telephone</td>\n",
       "      <td>may</td>\n",
       "      <td>mon</td>\n",
       "      <td>149</td>\n",
       "      <td>1</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.1</td>\n",
       "      <td>93.994</td>\n",
       "      <td>-36.4</td>\n",
       "      <td>4.857</td>\n",
       "      <td>5191.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>37</td>\n",
       "      <td>services</td>\n",
       "      <td>married</td>\n",
       "      <td>high.school</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>telephone</td>\n",
       "      <td>may</td>\n",
       "      <td>mon</td>\n",
       "      <td>226</td>\n",
       "      <td>1</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.1</td>\n",
       "      <td>93.994</td>\n",
       "      <td>-36.4</td>\n",
       "      <td>4.857</td>\n",
       "      <td>5191.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>40</td>\n",
       "      <td>admin.</td>\n",
       "      <td>married</td>\n",
       "      <td>basic.6y</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>telephone</td>\n",
       "      <td>may</td>\n",
       "      <td>mon</td>\n",
       "      <td>151</td>\n",
       "      <td>1</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.1</td>\n",
       "      <td>93.994</td>\n",
       "      <td>-36.4</td>\n",
       "      <td>4.857</td>\n",
       "      <td>5191.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>56</td>\n",
       "      <td>services</td>\n",
       "      <td>married</td>\n",
       "      <td>high.school</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>telephone</td>\n",
       "      <td>may</td>\n",
       "      <td>mon</td>\n",
       "      <td>307</td>\n",
       "      <td>1</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.1</td>\n",
       "      <td>93.994</td>\n",
       "      <td>-36.4</td>\n",
       "      <td>4.857</td>\n",
       "      <td>5191.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   age        job  marital    education  default housing loan    contact  \\\n",
       "0   56  housemaid  married     basic.4y       no      no   no  telephone   \n",
       "1   57   services  married  high.school  unknown      no   no  telephone   \n",
       "2   37   services  married  high.school       no     yes   no  telephone   \n",
       "3   40     admin.  married     basic.6y       no      no   no  telephone   \n",
       "4   56   services  married  high.school       no      no  yes  telephone   \n",
       "\n",
       "  month day_of_week  duration  campaign  pdays  previous     poutcome  \\\n",
       "0   may         mon       261         1    999         0  nonexistent   \n",
       "1   may         mon       149         1    999         0  nonexistent   \n",
       "2   may         mon       226         1    999         0  nonexistent   \n",
       "3   may         mon       151         1    999         0  nonexistent   \n",
       "4   may         mon       307         1    999         0  nonexistent   \n",
       "\n",
       "   emp.var.rate  cons.price.idx  cons.conf.idx  euribor3m  nr.employed  \n",
       "0           1.1          93.994          -36.4      4.857       5191.0  \n",
       "1           1.1          93.994          -36.4      4.857       5191.0  \n",
       "2           1.1          93.994          -36.4      4.857       5191.0  \n",
       "3           1.1          93.994          -36.4      4.857       5191.0  \n",
       "4           1.1          93.994          -36.4      4.857       5191.0  "
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data_original   = pd.read_csv('data/raw/bank-additional/bank-additional-full.csv', sep = \";\")\n",
    "data = data_original.copy()\n",
    "\n",
    "labels = data[\"y\"].copy()\n",
    "data   = data.drop(\"y\",axis=1)\n",
    "\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "# 1) Data splitting"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "To make sure we're testing our process as objectively, we don't want the data we use to fit the model (\"training\" data) to be mixed with the data used for the final performance test (\"testing\" data). \n",
    "\n",
    "First we'll try a simple 70/30, training/testing spit, randomly selecting data points for each set.\n",
    "\n",
    "We won't take any looks at the the data until it's split, so that the test can be as independent as possible from our analysis strategy. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "train, test, train_labels, test_labels = train_test_split(data, labels, test_size=0.33, random_state=1337)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's see what the data types are, and how many entries we have after the split."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 27595 entries, 14401 to 3223\n",
      "Data columns (total 20 columns):\n",
      "age               27595 non-null int64\n",
      "job               27595 non-null object\n",
      "marital           27595 non-null object\n",
      "education         27595 non-null object\n",
      "default           27595 non-null object\n",
      "housing           27595 non-null object\n",
      "loan              27595 non-null object\n",
      "contact           27595 non-null object\n",
      "month             27595 non-null object\n",
      "day_of_week       27595 non-null object\n",
      "duration          27595 non-null int64\n",
      "campaign          27595 non-null int64\n",
      "pdays             27595 non-null int64\n",
      "previous          27595 non-null int64\n",
      "poutcome          27595 non-null object\n",
      "emp.var.rate      27595 non-null float64\n",
      "cons.price.idx    27595 non-null float64\n",
      "cons.conf.idx     27595 non-null float64\n",
      "euribor3m         27595 non-null float64\n",
      "nr.employed       27595 non-null float64\n",
      "dtypes: float64(5), int64(5), object(10)\n",
      "memory usage: 4.4+ MB\n"
     ]
    }
   ],
   "source": [
    "train.info()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "So we have ourselves a mixed-bag of data here. We won't be able to just charge-forward with either a categorical or quantitative approach... \n",
    "\n",
    "Now let's see a sample of the data table itself."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>age</th>\n",
       "      <th>job</th>\n",
       "      <th>marital</th>\n",
       "      <th>education</th>\n",
       "      <th>default</th>\n",
       "      <th>housing</th>\n",
       "      <th>loan</th>\n",
       "      <th>contact</th>\n",
       "      <th>month</th>\n",
       "      <th>day_of_week</th>\n",
       "      <th>duration</th>\n",
       "      <th>campaign</th>\n",
       "      <th>pdays</th>\n",
       "      <th>previous</th>\n",
       "      <th>poutcome</th>\n",
       "      <th>emp.var.rate</th>\n",
       "      <th>cons.price.idx</th>\n",
       "      <th>cons.conf.idx</th>\n",
       "      <th>euribor3m</th>\n",
       "      <th>nr.employed</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>14401</th>\n",
       "      <td>38</td>\n",
       "      <td>blue-collar</td>\n",
       "      <td>single</td>\n",
       "      <td>basic.9y</td>\n",
       "      <td>unknown</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>jul</td>\n",
       "      <td>mon</td>\n",
       "      <td>72</td>\n",
       "      <td>3</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.4</td>\n",
       "      <td>93.918</td>\n",
       "      <td>-42.7</td>\n",
       "      <td>4.962</td>\n",
       "      <td>5228.1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35284</th>\n",
       "      <td>39</td>\n",
       "      <td>admin.</td>\n",
       "      <td>married</td>\n",
       "      <td>university.degree</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>may</td>\n",
       "      <td>fri</td>\n",
       "      <td>364</td>\n",
       "      <td>2</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>-1.8</td>\n",
       "      <td>92.893</td>\n",
       "      <td>-46.2</td>\n",
       "      <td>1.250</td>\n",
       "      <td>5099.1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1225</th>\n",
       "      <td>36</td>\n",
       "      <td>blue-collar</td>\n",
       "      <td>married</td>\n",
       "      <td>basic.6y</td>\n",
       "      <td>unknown</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>telephone</td>\n",
       "      <td>may</td>\n",
       "      <td>thu</td>\n",
       "      <td>166</td>\n",
       "      <td>1</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.1</td>\n",
       "      <td>93.994</td>\n",
       "      <td>-36.4</td>\n",
       "      <td>4.855</td>\n",
       "      <td>5191.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8289</th>\n",
       "      <td>53</td>\n",
       "      <td>self-employed</td>\n",
       "      <td>married</td>\n",
       "      <td>university.degree</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>telephone</td>\n",
       "      <td>jun</td>\n",
       "      <td>tue</td>\n",
       "      <td>119</td>\n",
       "      <td>4</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.4</td>\n",
       "      <td>94.465</td>\n",
       "      <td>-41.8</td>\n",
       "      <td>4.864</td>\n",
       "      <td>5228.1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31412</th>\n",
       "      <td>32</td>\n",
       "      <td>technician</td>\n",
       "      <td>single</td>\n",
       "      <td>professional.course</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>may</td>\n",
       "      <td>wed</td>\n",
       "      <td>97</td>\n",
       "      <td>2</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>-1.8</td>\n",
       "      <td>92.893</td>\n",
       "      <td>-46.2</td>\n",
       "      <td>1.334</td>\n",
       "      <td>5099.1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       age            job  marital            education  default housing loan  \\\n",
       "14401   38    blue-collar   single             basic.9y  unknown     yes   no   \n",
       "35284   39         admin.  married    university.degree       no     yes   no   \n",
       "1225    36    blue-collar  married             basic.6y  unknown      no   no   \n",
       "8289    53  self-employed  married    university.degree       no     yes   no   \n",
       "31412   32     technician   single  professional.course       no     yes   no   \n",
       "\n",
       "         contact month day_of_week  duration  campaign  pdays  previous  \\\n",
       "14401   cellular   jul         mon        72         3    999         0   \n",
       "35284   cellular   may         fri       364         2    999         0   \n",
       "1225   telephone   may         thu       166         1    999         0   \n",
       "8289   telephone   jun         tue       119         4    999         0   \n",
       "31412   cellular   may         wed        97         2    999         0   \n",
       "\n",
       "          poutcome  emp.var.rate  cons.price.idx  cons.conf.idx  euribor3m  \\\n",
       "14401  nonexistent           1.4          93.918          -42.7      4.962   \n",
       "35284  nonexistent          -1.8          92.893          -46.2      1.250   \n",
       "1225   nonexistent           1.1          93.994          -36.4      4.855   \n",
       "8289   nonexistent           1.4          94.465          -41.8      4.864   \n",
       "31412  nonexistent          -1.8          92.893          -46.2      1.334   \n",
       "\n",
       "       nr.employed  \n",
       "14401       5228.1  \n",
       "35284       5099.1  \n",
       "1225        5191.0  \n",
       "8289        5228.1  \n",
       "31412       5099.1  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Some columns show missing or N/A values, like 'unknown' in 'default', and 'nonexistent' in 'poutcome'. We may not necessarily keep all the columns, but we have to choose a strategy now for handling rows (\"participants\") that have missing data values.\n",
    "\n",
    "For this project, let's take a conservative approach- remove entire rows that have 'unknown' values. This is to avoid treating 'unknown' as carrying the same degree of meaning as the defined categories. We don't know why this data are unknown, and we can't assume that two people who have 'unknown' education values have anything in common. \n",
    "\n",
    "Value '999' for 'pdays' also carries no quantitative meaning. But at the same time, if we remove all of these rows, then we're removing all the data where the participant was being called for the first-time. We'll leave these rows for now, but consider removing the 'pdays' column later."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "14401    no\n",
       "35284    no\n",
       "1225     no\n",
       "8289     no\n",
       "31412    no\n",
       "Name: y, dtype: object"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_labels.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/work1/users/aaronb/Softbrary/Anaconda/lib/python2.7/site-packages/ipykernel_launcher.py:8: UserWarning: Boolean Series key will be reindexed to match DataFrame index.\n",
      "  \n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "35284    no\n",
       "8289     no\n",
       "31412    no\n",
       "22884    no\n",
       "25811    no\n",
       "Name: y, dtype: object"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def dropMissingData(data, labels, missing_val, missing_val_cols, labels_colname = \"y\"):\n",
    "    \n",
    "    data_temp = data.join(labels)\n",
    "    #print data_temp.head()\n",
    "    \n",
    "    for missing_val_col in missing_val_cols:\n",
    "        \n",
    "        data_temp = data_temp[ data[missing_val_col]!=missing_val  ]\n",
    "\n",
    "    labels_temp = data_temp[labels_colname]\n",
    "    data_temp   = data_temp.drop(labels_colname, axis=1)\n",
    "    \n",
    "    return data_temp, labels_temp\n",
    "\n",
    "train, train_labels = dropMissingData(train,\n",
    "                                      train_labels, \n",
    "                                      missing_val=\"unknown\", \n",
    "                                      missing_val_cols = [\"education\",\"marital\",\"default\",\"job\"],\n",
    "                                      labels_colname = \"y\")\n",
    "train_labels.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(train.default[train.default==\"unknown\"])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 3) Feature Selection"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We're given 21 features, which is tempting to just throw at a black-box and see what comes out.\n",
    "But let's show some patience, and pre-analyze which features are the most useful, which are redundant and uninformative."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In to the instructions, we're told that \"duration\" had best be omitted:  \n",
    "```...the duration is not known before a call is performed. Also, after the end of the call y is obviously known. Thus, this input should only be included for benchmark purposes and should be discarded if the intention is to have a realistic predictive model. ``` \n",
    "\n",
    "Which makes sense. If you are looking at a spread-sheet of client data, you have no idea how long a phone call in their future is going to last. If the call duration was being controlled somehow by the telemarketers, we could test if keeping participants on the phone longer related to more \"yes\" answers. But we don't know that in this case. We just assume that 'duration' means the after-the-fact length of the call, including time it takes for customers to ask questions and confirm their answers, and that the telemarketers were not artifically controlling the call length. \n",
    "\n",
    "So let's drop it."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "train = train.drop(\"duration\", axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3.2) Isolate the quantitative columns\n",
    "\n",
    "Some data are given to us as numbers. Others are string labels, without an obvious way to rank them."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The quantitative data are either integer ('int64') or floats ('float64'). The qualitative, string data are given as object data type, indicatd by 'O'. We'll call the qualitative ones 'nom_cols', so we don't confuse it with 'quant_cols'."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([dtype('int64'), dtype('O'), dtype('O'), dtype('O'), dtype('O'),\n",
       "       dtype('O'), dtype('O'), dtype('O'), dtype('O'), dtype('O'),\n",
       "       dtype('int64'), dtype('int64'), dtype('int64'), dtype('O'),\n",
       "       dtype('float64'), dtype('float64'), dtype('float64'),\n",
       "       dtype('float64'), dtype('float64')], dtype=object)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.dtypes.values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['age', 'campaign', 'pdays', 'previous', 'emp.var.rate',\n",
       "       'cons.price.idx', 'cons.conf.idx', 'euribor3m', 'nr.employed'], dtype=object)"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "quant_cols  = train.columns.values[(train.dtypes =='int64') | (train.dtypes =='float64')]\n",
    "quant_cols"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3.3) Variance test the quantitative features"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Asking \"How informative is this column?\" is like asking \"How much does this column tell us about how data points vary from one another?\". \n",
    "\n",
    "For the quantitative data, there's a straight-forward way to test this--- look at the variance. To do this properly, we have to put the data into the same scale. Depending on the modeling method we try, this might have to do this anyways (i.e. for neural networks).\n",
    "\n",
    "\n",
    "We'll first set a threshold for the variance, let's say 0.2, and then exclude columns that don't meet this threshold. A good example from scikit learn's documentations shows a variance filter of 0.8, so let's use that for the moment. (Important to keep in mind that the y-columns itself shows low variance, with the vast majority being \"No\"s.)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of 'Yes' responses: 2672\n",
      "Number of 'No' responses: 18325\n"
     ]
    }
   ],
   "source": [
    "n_yes = len(train_labels[train_labels==\"yes\"])\n",
    "n_no  = len(train_labels[train_labels==\"no\"])\n",
    "yes_no_ratio =  (n_yes/(n_yes+n_no))*100\n",
    "print \"Number of 'Yes' responses: {}\".format(n_yes)\n",
    "print \"Number of 'No' responses: {}\".format(n_no)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[  1.05763219e+02   7.23983410e+00   4.03346286e+04   2.70406619e-01\n",
      "   2.60177040e+00   3.42982542e-01   2.29885910e+01   3.16484341e+00\n",
      "   5.66210077e+03]\n",
      "[ True  True  True  True  True  True  True  True  True]\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>age</th>\n",
       "      <th>campaign</th>\n",
       "      <th>pdays</th>\n",
       "      <th>previous</th>\n",
       "      <th>emp.var.rate</th>\n",
       "      <th>cons.price.idx</th>\n",
       "      <th>cons.conf.idx</th>\n",
       "      <th>euribor3m</th>\n",
       "      <th>nr.employed</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>-0.01786</td>\n",
       "      <td>-0.522313</td>\n",
       "      <td>42.426204</td>\n",
       "      <td>-0.193266</td>\n",
       "      <td>-1.725975</td>\n",
       "      <td>-0.630254</td>\n",
       "      <td>-5.602115</td>\n",
       "      <td>-2.206138</td>\n",
       "      <td>-61.49658</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>13.98214</td>\n",
       "      <td>1.477687</td>\n",
       "      <td>42.426204</td>\n",
       "      <td>-0.193266</td>\n",
       "      <td>1.474025</td>\n",
       "      <td>0.941746</td>\n",
       "      <td>-1.202115</td>\n",
       "      <td>1.407862</td>\n",
       "      <td>67.50342</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>-7.01786</td>\n",
       "      <td>-0.522313</td>\n",
       "      <td>42.426204</td>\n",
       "      <td>-0.193266</td>\n",
       "      <td>-1.725975</td>\n",
       "      <td>-0.630254</td>\n",
       "      <td>-5.602115</td>\n",
       "      <td>-2.122138</td>\n",
       "      <td>-61.49658</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>-9.01786</td>\n",
       "      <td>0.477687</td>\n",
       "      <td>42.426204</td>\n",
       "      <td>-0.193266</td>\n",
       "      <td>1.474025</td>\n",
       "      <td>-0.079254</td>\n",
       "      <td>4.497885</td>\n",
       "      <td>1.508862</td>\n",
       "      <td>67.50342</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>-1.01786</td>\n",
       "      <td>-1.522313</td>\n",
       "      <td>42.426204</td>\n",
       "      <td>-0.193266</td>\n",
       "      <td>-0.025975</td>\n",
       "      <td>-0.323254</td>\n",
       "      <td>-1.402115</td>\n",
       "      <td>0.663862</td>\n",
       "      <td>35.20342</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        age  campaign      pdays  previous  emp.var.rate  cons.price.idx  \\\n",
       "0  -0.01786 -0.522313  42.426204 -0.193266     -1.725975       -0.630254   \n",
       "1  13.98214  1.477687  42.426204 -0.193266      1.474025        0.941746   \n",
       "2  -7.01786 -0.522313  42.426204 -0.193266     -1.725975       -0.630254   \n",
       "3  -9.01786  0.477687  42.426204 -0.193266      1.474025       -0.079254   \n",
       "4  -1.01786 -1.522313  42.426204 -0.193266     -0.025975       -0.323254   \n",
       "\n",
       "   cons.conf.idx  euribor3m  nr.employed  \n",
       "0      -5.602115  -2.206138    -61.49658  \n",
       "1      -1.202115   1.407862     67.50342  \n",
       "2      -5.602115  -2.122138    -61.49658  \n",
       "3       4.497885   1.508862     67.50342  \n",
       "4      -1.402115   0.663862     35.20342  "
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.feature_selection import VarianceThreshold\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "sel = VarianceThreshold(threshold=0.2)\n",
    "mmscale = StandardScaler(with_std=False)\n",
    "train_scaled = pd.DataFrame(\n",
    "    mmscale.fit_transform( train[quant_cols] ), \n",
    "    columns= train[quant_cols].columns)\n",
    "\n",
    "sel.fit_transform(train_scaled)\n",
    " \n",
    "print sel.variances_\n",
    "\n",
    "print sel.get_support()\n",
    "\n",
    "train_scaled.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Interestingly all of our quanttative features pass the variance test. Although two of them just barely, 'pdays' and 'previous'. These two must be correlated, since you can only have a previous outcome recorded if you have a valid value for pdays. In any case, the numeric value of \"999\" for pdays has no meaning."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "train = train.drop(['pdays','poutcome','previous','campaign'], axis=1)\n",
    "quant_cols  = train.columns.values[(train.dtypes =='int64') | (train.dtypes =='float64')]\n",
    "nom_cols = train.columns.values[train.dtypes == 'O']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 3.4) Check for redundance in the quantitative features"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's take the quantitative data aside for a minute, and see if any of the columns are storngly related or redundant. \n",
    "\n",
    "One way of doing this is through a simple correlation test. Pandas gives us a quick and convenient way to get a correlation coefficient (pearson, spearman) for all of the features in our dataset, vs. each other. Or something like a normalized covariance matrix.\n",
    "\n",
    "I prefer to use Spearman as the default corr. test, since it doesn't assume we're looking at linear trends, and is less sensitive to outliers."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_corrmat = train[quant_cols].corr(method=\"spearman\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x2af9d4a00e50>"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAF9CAYAAADoebhRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n+kIIYUKk9IUQwoRI6QshhAn5f0g3UtIt33CbAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2af9ce2658d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sb.heatmap(train_corrmat, annot=True,cbar=False, square=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here we can see something interesting: \n",
    "'euribor3m', 'nr.employed' and 'emp.var.rate' are strongly correlated.\n",
    "\n",
    "Per the data guide, these are all different ways of diagnosing the socio-economic environment, at the time of the call. They don't have anything to do with the participants themselves.\n",
    "\n",
    "Which one do we drop? While they are all \"quaterly\" average measures, only one of them gives a daily running average, and that's the 'euribor' (which affects the rate of return on term-deposits?). Perhaps this daily variation will tell us more about the specific conditions for each caller. This choice is arbitrary choice, but for now I want to simplify the dataset while developing a strategy.\n",
    "\n",
    "Also it should come as no surprise that a participants age has nothing to do with the economic conditions on the day of the call."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "redundant_cols = [\"nr.employed\", \"emp.var.rate\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "train = train.drop(redundant_cols, axis=1)\n",
    "quant_cols  = train.columns.values[(train.dtypes =='int64') | (train.dtypes =='float64')]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 4) Pre-analysis: Categorical data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now let's take a look at some of our categorical data. For starters, just look at t he attributes that are intrinsic attribut.\n",
    "es of the participant- not the phone they used, not what the economic weather was like. This leaves education, job, and age. Well, marital status, housing, financial status can arguably be \"intrinsic\", but let's start with the basics. \n",
    "\n",
    "Perhaps we can agree that age would be the most objective, quantitatve attribute of a caller, among the data we have. Let's start there:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "0\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfIAAAFYCAYAAACoFn5YAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n/Q+vUZK5vygkFgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2af9d49a2890>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Age distributions of yesses and nos:\n",
    "sb.violinplot(x=train_labels, y=train.age)\n",
    "print len(train_labels[(train_labels==\"yes\") & (train.age > 60)])/len(train_labels[train_labels==\"yes\"])\n",
    "       \n",
    "print len(train_labels[(train_labels==\"no\") & (train.age > 60)])/len(train_labels[train_labels==\"no\"])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Without getting to an objective measure here, it as at least visually obvious that the yesses (1s) have a thicker tail out past the retirement ages (though we can't assume yet that all these people are \"retired\" per-se)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x2af9d49c4dd0>"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAeoAAAFYCAYAAACcb79EAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngxHUAAAY7P8BjFmXKW3P1eAAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2af9de2defd0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sb.distplot(train.age, kde=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "When we look at the distribution of ages, we see that the \"violin\" plots from before can be a bit misleading.\n",
    "\n",
    "While they seem to tell us that there is an overall trend with age,the age distribution above below tells us that there are very few participants over age 60. So while it may be accurate to say that, within those over 60, there is a trend to say yes- there is likely not a significant trend in this data for those who say yes to be over 60. \n",
    "\n",
    "We can't rule out a samping bias, but we do have to be careful about density plots. The reason for the confusion is that, for each side of the violin plots, the area is normalized to that category. So it falseley represents the groups is being roughly equal."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfIAAAGCCAYAAAAbs0ERAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nf2tL7A6+1mCEAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2af9de2af2d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Education-age demographics\n",
    "g = sb.violinplot(x=train.education, y=train.age, hue=train_labels, split=True)\n",
    "tick_labels = g.set_xticklabels(g.get_xticklabels(),rotation=30)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfoAAAF2CAYAAACGZ1rqAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2af9de39e510>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Job-age demographic\n",
    "g = sb.violinplot(x=train.job, y=train.age, hue=train_labels, split=True)\n",
    "tick_labels = g.set_xticklabels(g.get_xticklabels(),rotation=30)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The \"retired\" category is interesting. Maybe we would not be wrong in assuming our retirement-aged yessers are in-fact retired. But then again, we see an extended advanced-age-tail with the \"housemaid\" category as well."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 5) Process the categorical data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "A few of our data columns lend themselves to straightforward numerical relabeling. For example, January -> 1, Monday -> 2. Not to say that these have an intrinsic quantitative meaning, but it makes sense to use the conventional integer labels rather than arbitrary ones, to maximixe readability."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Some easy things we can convert are month and day of the week to 1-12 and 1-7"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "def relabelData(input_labels, input_frame, column):\n",
    "    \n",
    "    for i in range(0,len(input_frame[column])):\n",
    "        for j in range(0,len(input_labels)):\n",
    "            if input_frame[column].values[i] == input_labels[j]:\n",
    "                input_frame[column].values[i] = j\n",
    "                \n",
    "    return input_frame[column]\n",
    "                \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "#%%timeit\n",
    "months = ['jan','feb','mar','apr','may','jun','jul','aug','sep','oct','nov','dec']  \n",
    "train.month = relabelData(months, train, \"month\")\n",
    "\n",
    "# Add one so that the month numbers correspond to conventional dates - Jan = 1, etc.\n",
    "train.month+=1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>age</th>\n",
       "      <th>job</th>\n",
       "      <th>marital</th>\n",
       "      <th>education</th>\n",
       "      <th>default</th>\n",
       "      <th>housing</th>\n",
       "      <th>loan</th>\n",
       "      <th>contact</th>\n",
       "      <th>month</th>\n",
       "      <th>day_of_week</th>\n",
       "      <th>cons.price.idx</th>\n",
       "      <th>cons.conf.idx</th>\n",
       "      <th>euribor3m</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>35284</th>\n",
       "      <td>39</td>\n",
       "      <td>admin.</td>\n",
       "      <td>married</td>\n",
       "      <td>university.degree</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>5</td>\n",
       "      <td>fri</td>\n",
       "      <td>92.893</td>\n",
       "      <td>-46.2</td>\n",
       "      <td>1.250</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8289</th>\n",
       "      <td>53</td>\n",
       "      <td>self-employed</td>\n",
       "      <td>married</td>\n",
       "      <td>university.degree</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>telephone</td>\n",
       "      <td>6</td>\n",
       "      <td>tue</td>\n",
       "      <td>94.465</td>\n",
       "      <td>-41.8</td>\n",
       "      <td>4.864</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31412</th>\n",
       "      <td>32</td>\n",
       "      <td>technician</td>\n",
       "      <td>single</td>\n",
       "      <td>professional.course</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>5</td>\n",
       "      <td>wed</td>\n",
       "      <td>92.893</td>\n",
       "      <td>-46.2</td>\n",
       "      <td>1.334</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22884</th>\n",
       "      <td>30</td>\n",
       "      <td>technician</td>\n",
       "      <td>divorced</td>\n",
       "      <td>high.school</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>8</td>\n",
       "      <td>mon</td>\n",
       "      <td>93.444</td>\n",
       "      <td>-36.1</td>\n",
       "      <td>4.965</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25811</th>\n",
       "      <td>38</td>\n",
       "      <td>blue-collar</td>\n",
       "      <td>married</td>\n",
       "      <td>basic.6y</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>11</td>\n",
       "      <td>wed</td>\n",
       "      <td>93.200</td>\n",
       "      <td>-42.0</td>\n",
       "      <td>4.120</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       age            job   marital            education default housing loan  \\\n",
       "35284   39         admin.   married    university.degree      no     yes   no   \n",
       "8289    53  self-employed   married    university.degree      no     yes   no   \n",
       "31412   32     technician    single  professional.course      no     yes   no   \n",
       "22884   30     technician  divorced          high.school      no      no   no   \n",
       "25811   38    blue-collar   married             basic.6y      no     yes   no   \n",
       "\n",
       "         contact month day_of_week  cons.price.idx  cons.conf.idx  euribor3m  \n",
       "35284   cellular     5         fri          92.893          -46.2      1.250  \n",
       "8289   telephone     6         tue          94.465          -41.8      4.864  \n",
       "31412   cellular     5         wed          92.893          -46.2      1.334  \n",
       "22884   cellular     8         mon          93.444          -36.1      4.965  \n",
       "25811   cellular    11         wed          93.200          -42.0      4.120  "
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Re-label days of the week"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>age</th>\n",
       "      <th>job</th>\n",
       "      <th>marital</th>\n",
       "      <th>education</th>\n",
       "      <th>default</th>\n",
       "      <th>housing</th>\n",
       "      <th>loan</th>\n",
       "      <th>contact</th>\n",
       "      <th>month</th>\n",
       "      <th>day_of_week</th>\n",
       "      <th>cons.price.idx</th>\n",
       "      <th>cons.conf.idx</th>\n",
       "      <th>euribor3m</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>35284</th>\n",
       "      <td>39</td>\n",
       "      <td>admin.</td>\n",
       "      <td>married</td>\n",
       "      <td>university.degree</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>5</td>\n",
       "      <td>6</td>\n",
       "      <td>92.893</td>\n",
       "      <td>-46.2</td>\n",
       "      <td>1.250</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8289</th>\n",
       "      <td>53</td>\n",
       "      <td>self-employed</td>\n",
       "      <td>married</td>\n",
       "      <td>university.degree</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>telephone</td>\n",
       "      <td>6</td>\n",
       "      <td>3</td>\n",
       "      <td>94.465</td>\n",
       "      <td>-41.8</td>\n",
       "      <td>4.864</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31412</th>\n",
       "      <td>32</td>\n",
       "      <td>technician</td>\n",
       "      <td>single</td>\n",
       "      <td>professional.course</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "      <td>92.893</td>\n",
       "      <td>-46.2</td>\n",
       "      <td>1.334</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22884</th>\n",
       "      <td>30</td>\n",
       "      <td>technician</td>\n",
       "      <td>divorced</td>\n",
       "      <td>high.school</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>8</td>\n",
       "      <td>2</td>\n",
       "      <td>93.444</td>\n",
       "      <td>-36.1</td>\n",
       "      <td>4.965</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25811</th>\n",
       "      <td>38</td>\n",
       "      <td>blue-collar</td>\n",
       "      <td>married</td>\n",
       "      <td>basic.6y</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>11</td>\n",
       "      <td>4</td>\n",
       "      <td>93.200</td>\n",
       "      <td>-42.0</td>\n",
       "      <td>4.120</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       age            job   marital            education default housing loan  \\\n",
       "35284   39         admin.   married    university.degree      no     yes   no   \n",
       "8289    53  self-employed   married    university.degree      no     yes   no   \n",
       "31412   32     technician    single  professional.course      no     yes   no   \n",
       "22884   30     technician  divorced          high.school      no      no   no   \n",
       "25811   38    blue-collar   married             basic.6y      no     yes   no   \n",
       "\n",
       "         contact month day_of_week  cons.price.idx  cons.conf.idx  euribor3m  \n",
       "35284   cellular     5           6          92.893          -46.2      1.250  \n",
       "8289   telephone     6           3          94.465          -41.8      4.864  \n",
       "31412   cellular     5           4          92.893          -46.2      1.334  \n",
       "22884   cellular     8           2          93.444          -36.1      4.965  \n",
       "25811   cellular    11           4          93.200          -42.0      4.120  "
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#%%timeit \n",
    "weekdays = ['sun','mon','tue','wed','thu','fri','sat']  \n",
    "train.day_of_week = relabelData(weekdays, train,\"day_of_week\")\n",
    "\n",
    "train.day_of_week+=1\n",
    "\n",
    "train.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "So those were fairly straightforward, since months and days of the week have an obvious order to them. \n",
    "\n",
    "But what do we do with things like \"job\" and \"education\"? We'll have to be a bit more arbitrary about it, if we want to do any numerical analysis or apply vectorized routines to speed things up. At least Education has a kind of chronology to it- high-school before university, and so on. \n",
    "\n",
    "If we enforce an order on something like \"job\" though, we're may be implicitly making assumptions about its correlation with other data. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The remaining columns don't have an obvious way to numberize them. We could convert education to years of education, but then what do we do with labels like 'illiterate'? We have to be careful though, not to transform the columns that we already dealt with."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['job', 'marital', 'education', 'default', 'housing', 'loan',\n",
       "       'contact'], dtype=object)"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "encode_cols = nom_cols[(nom_cols!=\"month\") & (nom_cols!=\"day_of_week\")].copy()\n",
    "encode_cols"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>job</th>\n",
       "      <th>marital</th>\n",
       "      <th>education</th>\n",
       "      <th>default</th>\n",
       "      <th>housing</th>\n",
       "      <th>loan</th>\n",
       "      <th>contact</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>35284</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>6</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8289</th>\n",
       "      <td>6</td>\n",
       "      <td>1</td>\n",
       "      <td>6</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31412</th>\n",
       "      <td>9</td>\n",
       "      <td>2</td>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22884</th>\n",
       "      <td>9</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25811</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       job  marital  education  default  housing  loan  contact\n",
       "35284    0        1          6        0        2     0        0\n",
       "8289     6        1          6        0        2     0        1\n",
       "31412    9        2          5        0        2     0        0\n",
       "22884    9        0          3        0        0     0        0\n",
       "25811    1        1          1        0        2     0        0"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn import preprocessing\n",
    "nom_encoder = preprocessing.LabelEncoder()\n",
    "train[encode_cols].apply(nom_encoder.fit_transform).head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>age</th>\n",
       "      <th>job</th>\n",
       "      <th>marital</th>\n",
       "      <th>education</th>\n",
       "      <th>default</th>\n",
       "      <th>housing</th>\n",
       "      <th>loan</th>\n",
       "      <th>contact</th>\n",
       "      <th>month</th>\n",
       "      <th>day_of_week</th>\n",
       "      <th>cons.price.idx</th>\n",
       "      <th>cons.conf.idx</th>\n",
       "      <th>euribor3m</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>35284</th>\n",
       "      <td>39</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>6</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>5</td>\n",
       "      <td>6</td>\n",
       "      <td>92.893</td>\n",
       "      <td>-46.2</td>\n",
       "      <td>1.250</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8289</th>\n",
       "      <td>53</td>\n",
       "      <td>6</td>\n",
       "      <td>1</td>\n",
       "      <td>6</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>6</td>\n",
       "      <td>3</td>\n",
       "      <td>94.465</td>\n",
       "      <td>-41.8</td>\n",
       "      <td>4.864</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31412</th>\n",
       "      <td>32</td>\n",
       "      <td>9</td>\n",
       "      <td>2</td>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "      <td>92.893</td>\n",
       "      <td>-46.2</td>\n",
       "      <td>1.334</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22884</th>\n",
       "      <td>30</td>\n",
       "      <td>9</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>8</td>\n",
       "      <td>2</td>\n",
       "      <td>93.444</td>\n",
       "      <td>-36.1</td>\n",
       "      <td>4.965</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25811</th>\n",
       "      <td>38</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>11</td>\n",
       "      <td>4</td>\n",
       "      <td>93.200</td>\n",
       "      <td>-42.0</td>\n",
       "      <td>4.120</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       age  job  marital  education  default  housing  loan  contact month  \\\n",
       "35284   39    0        1          6        0        2     0        0     5   \n",
       "8289    53    6        1          6        0        2     0        1     6   \n",
       "31412   32    9        2          5        0        2     0        0     5   \n",
       "22884   30    9        0          3        0        0     0        0     8   \n",
       "25811   38    1        1          1        0        2     0        0    11   \n",
       "\n",
       "      day_of_week  cons.price.idx  cons.conf.idx  euribor3m  \n",
       "35284           6          92.893          -46.2      1.250  \n",
       "8289            3          94.465          -41.8      4.864  \n",
       "31412           4          92.893          -46.2      1.334  \n",
       "22884           2          93.444          -36.1      4.965  \n",
       "25811           4          93.200          -42.0      4.120  "
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train[encode_cols] = train[encode_cols].apply(nom_encoder.fit_transform)\n",
    "train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "35284    no\n",
      "8289     no\n",
      "31412    no\n",
      "22884    no\n",
      "25811    no\n",
      "Name: y, dtype: object\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>y</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   y\n",
       "0  0\n",
       "1  0\n",
       "2  0\n",
       "3  0\n",
       "4  0"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "labels_encoder = preprocessing.LabelEncoder()\n",
    "\n",
    "print train_labels.head()\n",
    "#labels_encoder.fit_transform(train_labels)\n",
    "\n",
    "train_labels = pd.DataFrame(labels_encoder.fit_transform(train_labels), columns=[\"y\"])\n",
    "\n",
    "train_labels.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 6.1) Modelling experiments: Decision tree\n",
    "\n",
    "While I'm not an expert in decision trees-- I have only read about them -- among the methods I find for mixed-data type machine learning, I prefer it because of the relative transparency. A decision tree produces a graoh of its steps which we can easily visualize. MCA, while seemingly straightforward if we call it \"PCA for categorical data\", is more mathematically complex to describe. Neural nets are infamously opaque.\n",
    "\n",
    "What follows is largely based on scikit learn examples, either from their online documentation or \n",
    "http://scikit-learn.org/stable/modules/tree.html"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn import tree\n",
    "\n",
    "model = tree.DecisionTreeClassifier(random_state=1337) # We are using classifier rather than regressor, because or goal is to \"classify\" participants into \"yes\" and \"no\" categories.\n",
    "\n",
    "model = model.fit(train,train_labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/work1/users/aaronb/Softbrary/Anaconda/lib/python2.7/site-packages/ipykernel_launcher.py:8: UserWarning: Boolean Series key will be reindexed to match DataFrame index.\n",
      "  \n"
     ]
    }
   ],
   "source": [
    "def preProcess(data, labels, drop_cols, missing_val, missing_val_cols, labels_colname = \"y\"):\n",
    "    \n",
    "    from sklearn.preprocessing import LabelEncoder\n",
    "    \n",
    "    data = data.drop(drop_cols, axis=1)\n",
    "    \n",
    "    data, labels = dropMissingData(data, labels, missing_val, missing_val_cols, labels_colname = \"y\")\n",
    "    \n",
    "    nom_cols = data.columns.values[data.dtypes == 'O']    \n",
    "    \n",
    "    nom_encoder = LabelEncoder()\n",
    "\n",
    "    data[nom_cols] = data[nom_cols].apply(nom_encoder.fit_transform)\n",
    "    \n",
    "    labels_encoder = LabelEncoder()\n",
    "    \n",
    "    labels = pd.DataFrame(labels_encoder.fit_transform(labels), columns=[\"y\"])\n",
    "    \n",
    "    #print data.head()\n",
    "    \n",
    "    return data, labels\n",
    "\n",
    "\n",
    "drop_cols         = ['nr.employed', \n",
    "                     'emp.var.rate',\n",
    "                     'pdays',\n",
    "                     'poutcome',\n",
    "                     'previous',\n",
    "                     'campaign', \n",
    "                     'duration']\n",
    "\n",
    "missing_val       = 'unknown'\n",
    "\n",
    "missing_val_cols  = ['education','marital','job','default']\n",
    "    \n",
    "test, test_labels = preProcess(test, test_labels, drop_cols, missing_val, missing_val_cols)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Just for diagnostics, let's see how the model fits the training data. This is not a valid way to say if our model will work in reality. It just helps us see if something has gone terrible wrong so far. What we expect at this point is that the model works well for the training data. If it can't even fit data it has already seen however, that means we have other issues to resolve first."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "First, apply the model we just trained to the training data to get the predicted labels."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_predicted_labels = model.predict(train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Next we get a few performance metrics for the predictions."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy score: 0.986\n",
      "Precision score: 0.993\n",
      "Recall score: 0.893\n",
      "ROC score: 0.946\n",
      "f1 score: 0.94\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import accuracy_score, confusion_matrix, precision_score, recall_score, roc_auc_score, f1_score\n",
    "\n",
    "\n",
    "def reportPerformance(test_labels, predict_labels):\n",
    "    \n",
    "    print \"Accuracy score: {}\".format(  round( \n",
    "        accuracy_score(test_labels, predict_labels)\n",
    "        ,3))\n",
    "\n",
    "    print \"Precision score: {}\".format( round( \n",
    "        precision_score(test_labels, predict_labels)\n",
    "        ,3))\n",
    "\n",
    "    print \"Recall score: {}\".format(    round( \n",
    "        recall_score(test_labels, predict_labels)\n",
    "        ,3))\n",
    "\n",
    "    print \"ROC score: {}\".format(   round( \n",
    "        roc_auc_score(test_labels, predict_labels)\n",
    "        ,3))\n",
    "    \n",
    "    print \"f1 score: {}\".format(   round( \n",
    "        f1_score(test_labels, predict_labels, pos_label=1)\n",
    "        ,3))\n",
    "    \n",
    "reportPerformance(train_labels, train_predicted_labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The model performs suspiciously well on the training data...but of course it performs well on data it's already seen. The very high accuracy rate smells a bit of over fitting. Next we'll see how it does with the test data, which it's never encountered.\n",
    "\n",
    "To test our fitted model, we have to apply all of the above pre-processing steps to the test data as well. Let's bundle those into a function that we can use for later exeriments."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We'll also want to have a quick way to check various performanc measure of the model: Accuracy, precision, and recall. To be honest, accuracy is not really helpful in this case, since the main goal is finding the minority of cases where the result was \"yes\". \n",
    "\n",
    "A model that always predicts \"no\" would give us a misleadingly good accuracy.\n",
    "\n",
    "A model that always predicts \"yes\" (looking under every stone) would give us a perfect recall score, but in practice would waste hours of telemarketing time and resources.\n",
    "\n",
    "A 50/50 coin-flip model would give is a recall of about 50%. Same for accuracy. So we have to be careful to compare the scores together, and remember that it's entirely possible for a complex model to be less effective than asking your cat for advice."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "It seems our model doesn't generalize very well to data it hasn't seen yet. Perhaps the tree is being over-fit to the data?\n",
    "\n",
    "The recall rate of 0.333 says that we're recovering 1/3 of would-be depositors. Worse than flipping a coin.\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 6.2) Modelling experiments: Random forests\n",
    "\n",
    "\n",
    "Next we'll try an extension of the decision tree model, called \"random forests\". This is a way of preventing overfitting to the data, but constructing many different decision trees, with different subsets of the provided features and datasets.\n",
    "\n",
    "It's a bit more difficult to deal with though, since we have to choose the number of different trees generated in the procedure. The following cell fits the model for different values of 'n_estimators', the number of decision trees, and check which performs best via the OOB error rate."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/work1/users/aaronb/Softbrary/Anaconda/lib/python2.7/site-packages/ipykernel_launcher.py:20: DataConversionWarning: A column-vector y was passed when a 1d array was expected. Please change the shape of y to (n_samples,), for example using ravel().\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAgsAAAFaCAYAAACQZR4BAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2af9df33d7d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#Following example from stack exchange, for optimizing the number of estimators in the random forest\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from collections import OrderedDict\n",
    "\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "\n",
    "\n",
    "def findMinOOB(test, test_labels, min_estimators = 10, max_estimators = 1000, random_state = 1337):\n",
    "    \n",
    "    ensemble_clfs = [ (\"RandomForestClassifier, max_features=None\",RandomForestClassifier(random_state=random_state,warm_start=True, max_features=None, oob_score=True,))]\n",
    "\n",
    "\n",
    "    error_rate = OrderedDict((label, []) for label, _ in ensemble_clfs)\n",
    "\n",
    "    for label, clf in ensemble_clfs:\n",
    "        for i in range(min_estimators, max_estimators + 1):\n",
    "            clf.set_params(n_estimators=i)\n",
    "            clf.fit(test, test_labels)\n",
    "            oob_error = 1 - clf.oob_score_\n",
    "            error_rate[label].append((i, oob_error))\n",
    "\n",
    "\n",
    "    for label, clf_err in error_rate.items():\n",
    "        xs, ys = zip(*clf_err)\n",
    "        plt.plot(xs, ys, label=label)\n",
    "\n",
    "    plt.xlim(min_estimators, max_estimators)\n",
    "    plt.xlabel(\"n_estimators\")\n",
    "    plt.ylabel(\"OOB error rate\")\n",
    "    plt.legend(loc=\"upper right\")\n",
    "    plt.show()\n",
    "    \n",
    "    #print \"Min. OOB of {} found at N_estimators = {}\".format()\n",
    "    \n",
    "findMinOOB(test, test_labels, min_estimators = 100, max_estimators = 1000, random_state = 1337)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "While we see a steady decrease past 900, it's interesting that the minimum occurs at about 110. I'll prefer the more consistently lower error rate ~1000 for the moment. In a production environment though, it would probably be better to more thoroughly balance 'n_estimators' with the error rate, so that we don't waste computation time while re-testing."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "# from sklearn.model_selection import GridSearchCV\n",
    "\n",
    "# def getRandomForestBestParams(param_grid, data, labels, scoring = 'f1', random_state=1337):\n",
    "    \n",
    "#     estimator = RandomForestClassifier(random_state=random_state)\n",
    "\n",
    "#     gs = GridSearchCV(estimator, \n",
    "#                  param_grid,\n",
    "#                  scoring = 'f1')\n",
    "\n",
    "#     gs = gs.fit(train, train_labels.y)\n",
    "\n",
    "#     print \"Best Parameters: {}\".format(gs.best_params_)\n",
    "\n",
    "#     print \"Best Score: {}\".format(gs.best_score_)\n",
    "    \n",
    "#     return gs\n",
    "\n",
    "# param_grid = [\n",
    "#       {'n_estimators':range(100,1000,25),'max_features': range(2,(len(data.columns)-1))},\n",
    "#      ]\n",
    "\n",
    "# gs = getRandomForestBestParams(param_grid, train, train_labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we'll feed the best parameters into the model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/work1/users/aaronb/Softbrary/Anaconda/lib/python2.7/site-packages/ipykernel_launcher.py:7: DataConversionWarning: A column-vector y was passed when a 1d array was expected. Please change the shape of y to (n_samples,), for example using ravel().\n",
      "  import sys\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy score: 0.872\n",
      "Precision score: 0.473\n",
      "Recall score: 0.223\n",
      "ROC score: 0.594\n",
      "f1 score: 0.303\n"
     ]
    }
   ],
   "source": [
    "def predictWithRandomForest(train, train_labels, test, test_labels, n_estimators,  random_state=1337):\n",
    "\n",
    "    model_forest = RandomForestClassifier(n_estimators=n_estimators,\n",
    "                                          random_state=random_state, \n",
    "                                           oob_score=True) # We are using classifier rather than regressor, because or goal is to \"classify\" participants into \"yes\" and \"no\" categories.\n",
    "\n",
    "    model_forest = model_forest.fit(train, train_labels)\n",
    "\n",
    "\n",
    "    predict_labels = model_forest.predict(test)\n",
    "\n",
    "    reportPerformance(test_labels, predict_labels)\n",
    "    \n",
    "    return model_forest\n",
    "    \n",
    "model_forest = predictWithRandomForest(train, train_labels, test, test_labels, n_estimators=1000, random_state=1337)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 6.2 Modeling experiments: Undersampling"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The random forest approach doesn't really seem to help us out here, relative to the simple decision tree approach.\n",
    "\n",
    "Another issue is that our number of \"no\"s is way higher than the number of \"yesses\". More than simply a classification problem, this is really a recall problem. We're hunting for needles in the haystack, rather than sorting red fish from blue fish.'\n",
    "\n",
    "Next we'll play with the method of data splitting, rather than the model itself. We'll try and weight our training data more towards the \"Yes\" case, by inflating the number of yesses to match the number of nos. For this attempt, we'll arbirarily force the yes-to-no ratio to 0.5."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "36548\n",
      "4640\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "3526"
      ]
     },
     "execution_count": 60,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print len(labels[labels==\"no\"])\n",
    "print len(labels[labels==\"yes\"])\n",
    "2*5289/3"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "So let's take 2/3 of 5289 as the test set. But we want to try and use equal proportions of yes and no cases, so let's also take (2/3)*5289 of samples for the no cases."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of 'No' Cases: 36548/41188\n",
      "Number of 'Yes' Cases: 4640/41188\n"
     ]
    }
   ],
   "source": [
    "size = labels.size\n",
    "\n",
    "split_ratio = 2/3 #proportion of data taken as training\n",
    "\n",
    "size_yes    = len(labels[labels==\"yes\"])\n",
    "size_no     = len(labels[labels==\"no\"])\n",
    "\n",
    "\n",
    "print \"Number of 'No' Cases: {}/{}\".format(size_no,size)\n",
    "print \"Number of 'Yes' Cases: {}/{}\".format(size_yes,size)\n",
    "\n",
    "#print \"Number of 'No' Training Cases \" \n",
    "train_size_yes = (size_yes*2)//3 \n",
    "train_size_no  = train_size_yes\n",
    "\n",
    "test_size_yes = size_yes - train_size_yes\n",
    "test_size_no  = size_no  - train_size_no\n",
    "\n",
    "train_yes, test_yes, train_labels_yes, test_labels_yes   = train_test_split(data[labels==\"yes\"], labels[labels==\"yes\"], test_size=test_size_yes, random_state=1337)\n",
    "train_no , test_no,  train_labels_no,  test_labels_no    = train_test_split(data[labels==\"no\"],  labels[labels==\"no\"],  test_size=test_size_no, random_state=1337)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>age</th>\n",
       "      <th>job</th>\n",
       "      <th>marital</th>\n",
       "      <th>education</th>\n",
       "      <th>default</th>\n",
       "      <th>housing</th>\n",
       "      <th>loan</th>\n",
       "      <th>contact</th>\n",
       "      <th>month</th>\n",
       "      <th>day_of_week</th>\n",
       "      <th>duration</th>\n",
       "      <th>campaign</th>\n",
       "      <th>pdays</th>\n",
       "      <th>previous</th>\n",
       "      <th>poutcome</th>\n",
       "      <th>emp.var.rate</th>\n",
       "      <th>cons.price.idx</th>\n",
       "      <th>cons.conf.idx</th>\n",
       "      <th>euribor3m</th>\n",
       "      <th>nr.employed</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>56</td>\n",
       "      <td>housemaid</td>\n",
       "      <td>married</td>\n",
       "      <td>basic.4y</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>telephone</td>\n",
       "      <td>may</td>\n",
       "      <td>mon</td>\n",
       "      <td>261</td>\n",
       "      <td>1</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.1</td>\n",
       "      <td>93.994</td>\n",
       "      <td>-36.4</td>\n",
       "      <td>4.857</td>\n",
       "      <td>5191.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>57</td>\n",
       "      <td>services</td>\n",
       "      <td>married</td>\n",
       "      <td>high.school</td>\n",
       "      <td>unknown</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>telephone</td>\n",
       "      <td>may</td>\n",
       "      <td>mon</td>\n",
       "      <td>149</td>\n",
       "      <td>1</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.1</td>\n",
       "      <td>93.994</td>\n",
       "      <td>-36.4</td>\n",
       "      <td>4.857</td>\n",
       "      <td>5191.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>37</td>\n",
       "      <td>services</td>\n",
       "      <td>married</td>\n",
       "      <td>high.school</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>telephone</td>\n",
       "      <td>may</td>\n",
       "      <td>mon</td>\n",
       "      <td>226</td>\n",
       "      <td>1</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.1</td>\n",
       "      <td>93.994</td>\n",
       "      <td>-36.4</td>\n",
       "      <td>4.857</td>\n",
       "      <td>5191.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>40</td>\n",
       "      <td>admin.</td>\n",
       "      <td>married</td>\n",
       "      <td>basic.6y</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>telephone</td>\n",
       "      <td>may</td>\n",
       "      <td>mon</td>\n",
       "      <td>151</td>\n",
       "      <td>1</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.1</td>\n",
       "      <td>93.994</td>\n",
       "      <td>-36.4</td>\n",
       "      <td>4.857</td>\n",
       "      <td>5191.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>56</td>\n",
       "      <td>services</td>\n",
       "      <td>married</td>\n",
       "      <td>high.school</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>telephone</td>\n",
       "      <td>may</td>\n",
       "      <td>mon</td>\n",
       "      <td>307</td>\n",
       "      <td>1</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.1</td>\n",
       "      <td>93.994</td>\n",
       "      <td>-36.4</td>\n",
       "      <td>4.857</td>\n",
       "      <td>5191.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   age        job  marital    education  default housing loan    contact  \\\n",
       "0   56  housemaid  married     basic.4y       no      no   no  telephone   \n",
       "1   57   services  married  high.school  unknown      no   no  telephone   \n",
       "2   37   services  married  high.school       no     yes   no  telephone   \n",
       "3   40     admin.  married     basic.6y       no      no   no  telephone   \n",
       "4   56   services  married  high.school       no      no  yes  telephone   \n",
       "\n",
       "  month day_of_week  duration  campaign  pdays  previous     poutcome  \\\n",
       "0   may         mon       261         1    999         0  nonexistent   \n",
       "1   may         mon       149         1    999         0  nonexistent   \n",
       "2   may         mon       226         1    999         0  nonexistent   \n",
       "3   may         mon       151         1    999         0  nonexistent   \n",
       "4   may         mon       307         1    999         0  nonexistent   \n",
       "\n",
       "   emp.var.rate  cons.price.idx  cons.conf.idx  euribor3m  nr.employed  \n",
       "0           1.1          93.994          -36.4      4.857       5191.0  \n",
       "1           1.1          93.994          -36.4      4.857       5191.0  \n",
       "2           1.1          93.994          -36.4      4.857       5191.0  \n",
       "3           1.1          93.994          -36.4      4.857       5191.0  \n",
       "4           1.1          93.994          -36.4      4.857       5191.0  "
      ]
     },
     "execution_count": 62,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>age</th>\n",
       "      <th>job</th>\n",
       "      <th>marital</th>\n",
       "      <th>education</th>\n",
       "      <th>default</th>\n",
       "      <th>housing</th>\n",
       "      <th>loan</th>\n",
       "      <th>contact</th>\n",
       "      <th>month</th>\n",
       "      <th>day_of_week</th>\n",
       "      <th>duration</th>\n",
       "      <th>campaign</th>\n",
       "      <th>pdays</th>\n",
       "      <th>previous</th>\n",
       "      <th>poutcome</th>\n",
       "      <th>emp.var.rate</th>\n",
       "      <th>cons.price.idx</th>\n",
       "      <th>cons.conf.idx</th>\n",
       "      <th>euribor3m</th>\n",
       "      <th>nr.employed</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>8866</th>\n",
       "      <td>28</td>\n",
       "      <td>technician</td>\n",
       "      <td>single</td>\n",
       "      <td>professional.course</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>telephone</td>\n",
       "      <td>jun</td>\n",
       "      <td>thu</td>\n",
       "      <td>508</td>\n",
       "      <td>2</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.4</td>\n",
       "      <td>94.465</td>\n",
       "      <td>-41.8</td>\n",
       "      <td>4.866</td>\n",
       "      <td>5228.1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20238</th>\n",
       "      <td>34</td>\n",
       "      <td>admin.</td>\n",
       "      <td>married</td>\n",
       "      <td>university.degree</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>aug</td>\n",
       "      <td>mon</td>\n",
       "      <td>1238</td>\n",
       "      <td>1</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.4</td>\n",
       "      <td>93.444</td>\n",
       "      <td>-36.1</td>\n",
       "      <td>4.965</td>\n",
       "      <td>5228.1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30253</th>\n",
       "      <td>48</td>\n",
       "      <td>technician</td>\n",
       "      <td>divorced</td>\n",
       "      <td>high.school</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>apr</td>\n",
       "      <td>thu</td>\n",
       "      <td>400</td>\n",
       "      <td>1</td>\n",
       "      <td>999</td>\n",
       "      <td>1</td>\n",
       "      <td>failure</td>\n",
       "      <td>-1.8</td>\n",
       "      <td>93.075</td>\n",
       "      <td>-47.1</td>\n",
       "      <td>1.365</td>\n",
       "      <td>5099.1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30441</th>\n",
       "      <td>39</td>\n",
       "      <td>admin.</td>\n",
       "      <td>married</td>\n",
       "      <td>high.school</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>may</td>\n",
       "      <td>mon</td>\n",
       "      <td>303</td>\n",
       "      <td>3</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>-1.8</td>\n",
       "      <td>92.893</td>\n",
       "      <td>-46.2</td>\n",
       "      <td>1.354</td>\n",
       "      <td>5099.1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31011</th>\n",
       "      <td>45</td>\n",
       "      <td>blue-collar</td>\n",
       "      <td>married</td>\n",
       "      <td>basic.9y</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>may</td>\n",
       "      <td>tue</td>\n",
       "      <td>672</td>\n",
       "      <td>2</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>-1.8</td>\n",
       "      <td>92.893</td>\n",
       "      <td>-46.2</td>\n",
       "      <td>1.344</td>\n",
       "      <td>5099.1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       age          job   marital            education default housing loan  \\\n",
       "8866    28   technician    single  professional.course      no      no   no   \n",
       "20238   34       admin.   married    university.degree      no      no   no   \n",
       "30253   48   technician  divorced          high.school      no     yes   no   \n",
       "30441   39       admin.   married          high.school      no     yes   no   \n",
       "31011   45  blue-collar   married             basic.9y      no     yes   no   \n",
       "\n",
       "         contact month day_of_week  duration  campaign  pdays  previous  \\\n",
       "8866   telephone   jun         thu       508         2    999         0   \n",
       "20238   cellular   aug         mon      1238         1    999         0   \n",
       "30253   cellular   apr         thu       400         1    999         1   \n",
       "30441   cellular   may         mon       303         3    999         0   \n",
       "31011   cellular   may         tue       672         2    999         0   \n",
       "\n",
       "          poutcome  emp.var.rate  cons.price.idx  cons.conf.idx  euribor3m  \\\n",
       "8866   nonexistent           1.4          94.465          -41.8      4.866   \n",
       "20238  nonexistent           1.4          93.444          -36.1      4.965   \n",
       "30253      failure          -1.8          93.075          -47.1      1.365   \n",
       "30441  nonexistent          -1.8          92.893          -46.2      1.354   \n",
       "31011  nonexistent          -1.8          92.893          -46.2      1.344   \n",
       "\n",
       "       nr.employed  \n",
       "8866        5228.1  \n",
       "20238       5228.1  \n",
       "30253       5099.1  \n",
       "30441       5099.1  \n",
       "31011       5099.1  "
      ]
     },
     "execution_count": 63,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_yes.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Shuffle the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.utils import shuffle\n",
    "\n",
    "def shuffleTogether(data_yes, data_no, labels_yes, labels_no, random_state = 1337, labels_col = \"y\"):\n",
    "    \n",
    "    data   = data_yes.append(data_no)\n",
    "    labels = labels_yes.append(labels_no)\n",
    "    \n",
    "    data = data.join(labels)\n",
    "    \n",
    "    data = shuffle(data, random_state = random_state)\n",
    "    \n",
    "    labels = data[labels_col]\n",
    "    \n",
    "    data   = data.drop(labels_col, axis=1)\n",
    "    \n",
    "    return data, labels\n",
    "\n",
    "train_under, train_labels_under = shuffleTogether(train_yes, train_no, train_labels_yes, train_labels_no)\n",
    "test_under, test_labels_under   = shuffleTogether(test_yes,  test_no, test_labels_yes, test_labels_no)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>age</th>\n",
       "      <th>job</th>\n",
       "      <th>marital</th>\n",
       "      <th>education</th>\n",
       "      <th>default</th>\n",
       "      <th>housing</th>\n",
       "      <th>loan</th>\n",
       "      <th>contact</th>\n",
       "      <th>month</th>\n",
       "      <th>day_of_week</th>\n",
       "      <th>duration</th>\n",
       "      <th>campaign</th>\n",
       "      <th>pdays</th>\n",
       "      <th>previous</th>\n",
       "      <th>poutcome</th>\n",
       "      <th>emp.var.rate</th>\n",
       "      <th>cons.price.idx</th>\n",
       "      <th>cons.conf.idx</th>\n",
       "      <th>euribor3m</th>\n",
       "      <th>nr.employed</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>8522</th>\n",
       "      <td>31</td>\n",
       "      <td>blue-collar</td>\n",
       "      <td>married</td>\n",
       "      <td>basic.6y</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>telephone</td>\n",
       "      <td>jun</td>\n",
       "      <td>wed</td>\n",
       "      <td>406</td>\n",
       "      <td>2</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.4</td>\n",
       "      <td>94.465</td>\n",
       "      <td>-41.8</td>\n",
       "      <td>4.864</td>\n",
       "      <td>5228.1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19967</th>\n",
       "      <td>50</td>\n",
       "      <td>blue-collar</td>\n",
       "      <td>married</td>\n",
       "      <td>basic.6y</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>aug</td>\n",
       "      <td>fri</td>\n",
       "      <td>733</td>\n",
       "      <td>3</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>1.4</td>\n",
       "      <td>93.444</td>\n",
       "      <td>-36.1</td>\n",
       "      <td>4.966</td>\n",
       "      <td>5228.1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39495</th>\n",
       "      <td>77</td>\n",
       "      <td>retired</td>\n",
       "      <td>married</td>\n",
       "      <td>high.school</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>apr</td>\n",
       "      <td>thu</td>\n",
       "      <td>165</td>\n",
       "      <td>7</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>-1.8</td>\n",
       "      <td>93.749</td>\n",
       "      <td>-34.6</td>\n",
       "      <td>0.644</td>\n",
       "      <td>5008.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37878</th>\n",
       "      <td>53</td>\n",
       "      <td>admin.</td>\n",
       "      <td>married</td>\n",
       "      <td>university.degree</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>aug</td>\n",
       "      <td>mon</td>\n",
       "      <td>267</td>\n",
       "      <td>1</td>\n",
       "      <td>999</td>\n",
       "      <td>1</td>\n",
       "      <td>failure</td>\n",
       "      <td>-2.9</td>\n",
       "      <td>92.201</td>\n",
       "      <td>-31.4</td>\n",
       "      <td>0.821</td>\n",
       "      <td>5076.2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29148</th>\n",
       "      <td>53</td>\n",
       "      <td>admin.</td>\n",
       "      <td>married</td>\n",
       "      <td>basic.6y</td>\n",
       "      <td>no</td>\n",
       "      <td>yes</td>\n",
       "      <td>no</td>\n",
       "      <td>cellular</td>\n",
       "      <td>apr</td>\n",
       "      <td>fri</td>\n",
       "      <td>100</td>\n",
       "      <td>1</td>\n",
       "      <td>999</td>\n",
       "      <td>0</td>\n",
       "      <td>nonexistent</td>\n",
       "      <td>-1.8</td>\n",
       "      <td>93.075</td>\n",
       "      <td>-47.1</td>\n",
       "      <td>1.405</td>\n",
       "      <td>5099.1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       age          job  marital          education default housing loan  \\\n",
       "8522    31  blue-collar  married           basic.6y      no     yes   no   \n",
       "19967   50  blue-collar  married           basic.6y      no      no   no   \n",
       "39495   77      retired  married        high.school      no     yes   no   \n",
       "37878   53       admin.  married  university.degree      no     yes   no   \n",
       "29148   53       admin.  married           basic.6y      no     yes   no   \n",
       "\n",
       "         contact month day_of_week  duration  campaign  pdays  previous  \\\n",
       "8522   telephone   jun         wed       406         2    999         0   \n",
       "19967   cellular   aug         fri       733         3    999         0   \n",
       "39495   cellular   apr         thu       165         7    999         0   \n",
       "37878   cellular   aug         mon       267         1    999         1   \n",
       "29148   cellular   apr         fri       100         1    999         0   \n",
       "\n",
       "          poutcome  emp.var.rate  cons.price.idx  cons.conf.idx  euribor3m  \\\n",
       "8522   nonexistent           1.4          94.465          -41.8      4.864   \n",
       "19967  nonexistent           1.4          93.444          -36.1      4.966   \n",
       "39495  nonexistent          -1.8          93.749          -34.6      0.644   \n",
       "37878      failure          -2.9          92.201          -31.4      0.821   \n",
       "29148  nonexistent          -1.8          93.075          -47.1      1.405   \n",
       "\n",
       "       nr.employed  \n",
       "8522        5228.1  \n",
       "19967       5228.1  \n",
       "39495       5008.7  \n",
       "37878       5076.2  \n",
       "29148       5099.1  "
      ]
     },
     "execution_count": 65,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_under.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Apply same pre-processing as before"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/work1/users/aaronb/Softbrary/Anaconda/lib/python2.7/site-packages/ipykernel_launcher.py:8: UserWarning: Boolean Series key will be reindexed to match DataFrame index.\n",
      "  \n"
     ]
    }
   ],
   "source": [
    "train_under, train_labels_under  = preProcess(train_under, train_labels_under, drop_cols, missing_val, missing_val_cols)\n",
    "test_under,  test_labels_under   = preProcess(test_under,  test_labels_under,  drop_cols, missing_val, missing_val_cols)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "findMinOOB(train_under, train_labels_under, min_estimators = 100, max_estimators = 1500, random_state = 1337)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/work1/users/aaronb/Softbrary/Anaconda/lib/python2.7/site-packages/ipykernel_launcher.py:7: DataConversionWarning: A column-vector y was passed when a 1d array was expected. Please change the shape of y to (n_samples,), for example using ravel().\n",
      "  import sys\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy score: 0.735\n",
      "Precision score: 0.125\n",
      "Recall score: 0.706\n",
      "ROC score: 0.721\n",
      "f1 score: 0.212\n"
     ]
    }
   ],
   "source": [
    "from sklearn.ensemble import RandomForestClassifier\n",
    "\n",
    "model_forest = RandomForestClassifier(n_estimators=500,\n",
    "                                      random_state=1337, \n",
    "                                      oob_score=True) # We are using classifier rather than regressor, because or goal is to \"classify\" participants into \"yes\" and \"no\" categories.\n",
    "\n",
    "model_forest = model_forest.fit(train_under, train_labels_under)\n",
    "\n",
    "\n",
    "predict_labels_under = model_forest.predict(test_under)\n",
    "\n",
    "reportPerformance(test_labels_under, predict_labels_under)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_forest.feature_importance"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "While we get the best recall rate among any of the methods here, our method is still very crude. It tends to recognize the \"Yes\" customers, sure, but it also gives us a lot of false-positives."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Conclusion:\n",
    "\n",
    "I have tried several modeling approaches to predict the outcome of a telemarketing campagin. Of all the approaches, the most effective \n",
    "\n",
    "I've avoided making a model that predicts based on the length of the calls, as this seems a bit circular. This differs from many examples I have found online, which found that call duration is a good predictor. My hope was to try and explore a bit deeper than this, and predict based on more independent data.\n",
    "\n",
    "The bottom line is that the results are not very positive. But why? Well, if I were to take more time to get to the bottom of it, I would explore more rigorous approaches to tuning the models. The other interesting thing, which appears in the article connected to this data, is that they made a reasonable prediction using neural networks. This is something I would like to try if I had more time, but so far I avoided it because I do not feel neural networks are a good \"default\" approach, unless the task is something like image classification. The reason is that it is hard to visualize what a NN has actually done, at the end, when we are dealing with a mix of quantitative and qualitative data. Especially socio economic data... but perhaps I'm being too pessimistic, and it's worth a try to see how NNs compare to random forest approaches, SVMs,  etc.\n",
    "\n",
    "The other point is data pre-processing. Although \"inflating\" the proportion of \"Yes\" samples in the training data still only resulted in 70% recall, and a lot of false-positives (translating to a lot of wasted time calling \"Nos\"), it's important to note that this did increase the recall relative to the random-sampling case (training Yes/No proportions equal to that of the full data set.) Interesting, but not shocking: we gave our model more Yes examples, so it tended to guess \"Yes\" more frequently when it saw the test data. While accuracy of ~75%, and recall of ~70% are better than a coin-flip, this is probably not enough of a trade-off to implement the model in actual callee-selection. Perhaps the undersampling method however could be tuned more for better results.\n",
    "\n",
    "Based on the simple pre-analysis I did, using violin plots, etc., I realize that there is one data dimension seriously lacking here. That is an objective, quantitative indicator of a customers income, or better yet, disposable income. After all, someone cannot sign up for a term-deposit if they do not have a surplus of cash to deposit. Even just the fact that someone is wealthy is not an indicator of this, since their budget could be strictly balanced. As of now, there is no data that separates the objective \"Ability to deposit\" from the subjective \"Willingness to deposit\".\n",
    "\n",
    "My main reason for suspecting the need for such data, is the relatively high percentage of \"Yes\" responses for customers over the age of 60, and the same for the \"retired\" category. While it is speculative, it may be the case that some of these are individuals with lifetime savings, open to wealth-management options. But by the same token, as the advanced age group overlaps with the low-eduation (basic.4yr) group, we could speculate that some of them did not understand the details of the call. It's worth considering, and maybe getting more data, but not nearly strong enough evidence to make a marketing recommendation."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 6.3***) Modeling experiments:  Playing with MCA\n",
    "\n",
    "* **This section was added just out of my own curiosity. It is basically me playing with a completely new method I'd just heard about. \n",
    "Feel free to skip it on just read the Conclusions section that follows**\n",
    "\n",
    "Another option, for dealing with categorical labels, is Multiple Correspondance Analysis. This seems to be something of a categorical analog of Principal Component Analysis. What follows is essentially an attempt to apply the steps in the Python MCA tutorial (http://nbviewer.jupyter.org/github/esafak/mca/blob/master/docs/mca-BurgundiesExample.ipynb) to the present dataset. They had originally applied MCA to wine-tasting data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {},
   "outputs": [],
   "source": [
    "import mca"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.set_printoptions(formatter={'float': '{: 0.4f}'.format})\n",
    "pd.set_option('display.precision', 5)\n",
    "pd.set_option('display.max_columns', 25)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {},
   "outputs": [
    {
     "ename": "TypeError",
     "evalue": "No loop matching the specified signature and casting\nwas found for ufunc svd_n_f",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mTypeError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-95-87ca82a55884>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0mncols\u001b[0m   \u001b[0;34m=\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtrain\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mnom_cols\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcolumns\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 2\u001b[0;31m \u001b[0mmca_ind\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmca\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mMCA\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtrain\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mnom_cols\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mncols\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mncols\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mbenzecri\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mFalse\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      3\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0;32mprint\u001b[0m \u001b[0mmca\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mMCA\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__doc__\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/home/aaronb/.local/lib/python2.7/site-packages/mca.pyc\u001b[0m in \u001b[0;36m__init__\u001b[0;34m(self, DF, cols, ncols, benzecri, TOL, sparse, approximate)\u001b[0m\n\u001b[1;32m     78\u001b[0m                 \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     79\u001b[0m                         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_numitems\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mDF\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 80\u001b[0;31m                         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mP\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0ms\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mQ\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0msvd\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mproduct\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     81\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     82\u001b[0m                 \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mE\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mNone\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/work1/users/aaronb/Softbrary/Anaconda/lib/python2.7/site-packages/numpy/linalg/linalg.pyc\u001b[0m in \u001b[0;36msvd\u001b[0;34m(a, full_matrices, compute_uv)\u001b[0m\n\u001b[1;32m   1387\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1388\u001b[0m         \u001b[0msignature\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m'D->DdD'\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0misComplexType\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mt\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32melse\u001b[0m \u001b[0;34m'd->ddd'\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1389\u001b[0;31m         \u001b[0mu\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0ms\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mvt\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mgufunc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0ma\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msignature\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0msignature\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mextobj\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mextobj\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1390\u001b[0m         \u001b[0mu\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mu\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mastype\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mresult_t\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcopy\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mFalse\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1391\u001b[0m         \u001b[0ms\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0ms\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mastype\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0m_realType\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mresult_t\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcopy\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mFalse\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mTypeError\u001b[0m: No loop matching the specified signature and casting\nwas found for ufunc svd_n_f"
     ]
    }
   ],
   "source": [
    "ncols   = len(train[nom_cols].columns)\n",
    "mca_ind = mca.MCA(train[nom_cols], ncols=ncols, benzecri=False)\n",
    "\n",
    "print mca.MCA.__doc__\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'mca_ind' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-98-5f9d43d7c525>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mmca_ind\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mL\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      2\u001b[0m \u001b[0mmca_ind\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mexpl_var\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mgreenacre\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mFalse\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mN\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m2\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mNameError\u001b[0m: name 'mca_ind' is not defined"
     ]
    }
   ],
   "source": [
    "mca_ind.L\n",
    "mca_ind.expl_var(greenacre=False, N=2)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = np.array([mca_ind.L[:2], \n",
    "                 mca_ind.expl_var(greenacre=False, N=2) * 100]).T\n",
    "df = pd.DataFrame(data=data, columns=['cλ','%c'], index=range(1,3))\n",
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = np.array([mca_ind.L[:2], \n",
    "                 mca_ind.expl_var(greenacre=False, N=2) * 100]).T\n",
    "df = pd.DataFrame(data=data, columns=['cλ','%c'], index=range(1,3))\n",
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "fs, cos, cont = 'Factor score','Squared cosines', 'Contributions x 1000'\n",
    "table3 = pd.DataFrame(columns=train[nom_cols].index, index=pd.MultiIndex\n",
    "                      .from_product([[fs, cos, cont], range(1, 3)]))\n",
    "\n",
    "table3.loc[fs,    :] = mca_ind.fs_r(N=2).T\n",
    "table3.loc[cos,   :] = mca_ind.cos_r(N=2).T\n",
    "table3.loc[cont,  :] = mca_ind.cont_r(N=2).T * 1000\n",
    "#table3.loc[fs, 'W?'] = mca_ben.fs_r_sup(pd.DataFrame([i_sup]), N=2)[0]\n",
    "\n",
    "np.round(table3.astype(float), 2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "table4 = pd.DataFrame(columns=train[nom_cols].columns, index=pd.MultiIndex\n",
    "                      .from_product([[fs, cos, cont], range(1, 3)]))\n",
    "table4.loc[fs,  :] = mca_ind.fs_c(N=2).T\n",
    "table4.loc[cos, :] = mca_ind.cos_c(N=2).T\n",
    "table4.loc[cont,:] = mca_ind.cont_c(N=2).T * 1000\n",
    "\n",
    "\n",
    "np.round(table4.astype(float), 2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "points = table3.loc[fs].values\n",
    "labels = table3.columns.values\n",
    "\n",
    "plt.figure()\n",
    "#plt.margins(0.1)\n",
    "plt.axhline(0, color='gray')\n",
    "plt.axvline(0, color='gray')\n",
    "plt.xlabel('Factor 1')\n",
    "plt.ylabel('Factor 2')\n",
    "plt.scatter(*points, s=120, marker='o', alpha=0.5, linewidths=0)\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can see that MCA pulls out a primary component of our data, explaining 73% of variance. But what we are not seeing is any kind of structure appearing in the data, such that we could easily apply a cluster-based approach like K-means.\n",
    "Why? Well, maybe it's simply the case that these intrinsic properties of the participants- education job, etc., are not particularly informative when it comes to success rate for the telemarketing campaigns. \n",
    "\n",
    "There's one last thing we can try, which is to perform MCA with the labels as an input factor"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 8) Reference\n",
    "\n",
    "**This is just a copy of the instructions provided on the data archive page**\n",
    "\n",
    "The webpage for the dataset is located here: https://archive.ics.uci.edu/ml/datasets/bank+marketing\n",
    "\n",
    "Simple descriptions for each column are given as follows:\n",
    "Input variables:\n",
    "# bank client data:\n",
    "1 - age (numeric)  \n",
    "2 - job : type of job (categorical: 'admin.','blue-collar','entrepreneur','housemaid','management','retired','self-employed','services','student','technician','unemployed','unknown')  \n",
    "3 - marital : marital status (categorical: 'divorced','married','single','unknown'; note: 'divorced' means divorced or widowed)  \n",
    "4 - education (categorical: 'basic.4y','basic.6y','basic.9y','high.school','illiterate','professional.course','university.degree','unknown')  \n",
    "5 - default: has credit in default? (categorical: 'no','yes','unknown')  \n",
    "6 - housing: has housing loan? (categorical: 'no','yes','unknown')  \n",
    "7 - loan: has personal loan? (categorical: 'no','yes','unknown')  \n",
    "# related with the last contact of the current campaign:\n",
    "8 - contact: contact communication type (categorical: 'cellular','telephone')   \n",
    "9 - month: last contact month of year (categorical: 'jan', 'feb', 'mar', ..., 'nov', 'dec')  \n",
    "10 - day_of_week: last contact day of the week (categorical: 'mon','tue','wed','thu','fri')  \n",
    "11 - duration: last contact duration, in seconds (numeric). Important note: this attribute highly affects the output target (e.g., if duration=0 then y='no').   \n",
    "Yet, the duration is not known before a call is performed. Also, after the end of the call y is obviously known. Thus, this input should only be included for benchmark purposes and should be discarded if the intention is to have a realistic predictive model.  \n",
    "# other attributes:\n",
    "12 - campaign: number of contacts performed during this campaign and for this client (numeric, includes last contact)  \n",
    "13 - pdays: number of days that passed by after the client was last contacted from a previous campaign (numeric; 999 means client was not previously contacted)  \n",
    "14 - previous: number of contacts performed before this campaign and for this client (numeric)  \n",
    "15 - poutcome: outcome of the previous marketing campaign (categorical: 'failure','nonexistent','success')  \n",
    "# social and economic context attributes\n",
    "16 - emp.var.rate: employment variation rate - quarterly indicator (numeric)  \n",
    "17 - cons.price.idx: consumer price index - monthly indicator (numeric)   \n",
    "18 - cons.conf.idx: consumer confidence index - monthly indicator (numeric)   \n",
    "19 - euribor3m: euribor 3 month rate - daily indicator (numeric)  \n",
    "20 - nr.employed: number of employees - quarterly indicator (numeric)  \n",
    "  \n",
    "Output variable (desired target):  \n",
    "21 - y - has the client subscribed a term deposit? (binary: 'yes','no')  "
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.12"
  },
  "latex_envs": {
   "LaTeX_envs_menu_present": true,
   "autocomplete": true,
   "bibliofile": "reference.bib",
   "cite_by": "apalike",
   "current_citInitial": 1,
   "eqLabelWithNumbers": true,
   "eqNumInitial": 1,
   "hotkeys": {
    "equation": "Ctrl-E",
    "itemize": "Ctrl-I"
   },
   "labels_anchors": false,
   "latex_user_defs": false,
   "report_style_numbering": true,
   "user_envs_cfg": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}