SMB notebook-ES.ipynb

{
 "metadata": {
  "name": "",
  "signature": "sha256:56d152eb0c9db79b2d3b110c3b61f353f9afa60166165ca1c0b5d69a230cca71"
 },
 "nbformat": 3,
 "nbformat_minor": 0,
 "worksheets": [
  {
   "cells": [
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "Predicting Business Data for SMB\n",
      "================================"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "%matplotlib inline\n",
      "%load_ext autoreload\n",
      "%autoreload 2"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "The autoreload extension is already loaded. To reload it, use:\n",
        "  %reload_ext autoreload\n"
       ]
      }
     ],
     "prompt_number": 122
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "import pandas as pd\n",
      "from sklearn import linear_model, cross_validation\n",
      "from smb_insights.schedules import *\n",
      "from smb_insights.service_point_selector import ServicePointSelector"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 123
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# qh = load_quarter_hour_data()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 124
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "qh_d = load_quarter_hour_data_d()\n",
      "# qh_d = load_quarter_hour_data_d_july()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 125
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "qh_r = load_quarter_hour_data_r()\n",
      "# qh_r = load_quarter_hour_data_r_july()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 126
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# load school data\n",
      "# qh_s = load_quarter_hour_data_s()\n",
      "# qh_s_jan = load_quarter_hour_data_s_jan()\n",
      "# qh_s = load_quarter_hour_data_s_jan()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 127
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# id = 0\n",
      "# plt.subplot(3,1,1)\n",
      "# plt.plot(qh_r.iloc[id][usage_columns])\n",
      "# plt.yticks(np.linspace(0,.2,5))\n",
      "# plt.tick_params(\\\n",
      "#     axis='x',          # changes apply to the x-axis\n",
      "#     which='both',      # both major and minor ticks are affected\n",
      "#     bottom='off',      # ticks along the bottom edge are off\n",
      "#     top='off',         # ticks along the top edge are off\n",
      "#     labelbottom='off') \n",
      "# plt.subplot(3,1,2)\n",
      "# plt.plot(qh_r.iloc[id+1][usage_columns])\n",
      "# plt.yticks(np.linspace(0,.2,5))\n",
      "# plt.tick_params(\\\n",
      "#     axis='x',          # changes apply to the x-axis\n",
      "#     which='both',      # both major and minor ticks are affected\n",
      "#     bottom='off',      # ticks along the bottom edge are off\n",
      "#     top='off',         # ticks along the top edge are off\n",
      "#     labelbottom='off') \n",
      "# util_id = qh_r.iloc[id]['util_service_point_id']\n",
      "# print util_id\n",
      "# plt.subplot(3,1,3)\n",
      "# plt.plot(qh_r[qh_r['util_service_point_id']==util_id][usage_columns].mean())\n",
      "# plt.yticks(np.linspace(0,.2,5))\n",
      "# plt.xticks(range(0, 96, 16), times_of_day()[::16])"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 128
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# max_val = 0.75\n",
      "# id = 0\n",
      "# plt.subplot(3,1,1)\n",
      "# plt.plot(qh_d.iloc[id][usage_columns])\n",
      "# plt.yticks(np.linspace(0,max_val,5))\n",
      "# plt.tick_params(\\\n",
      "#     axis='x',          # changes apply to the x-axis\n",
      "#     which='both',      # both major and minor ticks are affected\n",
      "#     bottom='off',      # ticks along the bottom edge are off\n",
      "#     top='off',         # ticks along the top edge are off\n",
      "#     labelbottom='off') \n",
      "# plt.subplot(3,1,2)\n",
      "# plt.plot(qh_d.iloc[id+1][usage_columns])\n",
      "# plt.yticks(np.linspace(0,max_val,5))\n",
      "# plt.tick_params(\\\n",
      "#     axis='x',          # changes apply to the x-axis\n",
      "#     which='both',      # both major and minor ticks are affected\n",
      "#     bottom='off',      # ticks along the bottom edge are off\n",
      "#     top='off',         # ticks along the top edge are off\n",
      "#     labelbottom='off') \n",
      "# util_id = qh_d.iloc[id]['util_service_point_id']\n",
      "# print util_id\n",
      "# #plt.plot(qh_d.iloc[2][usage_columns])\n",
      "# #plt.savefig('foo_%s_%d.png' % (\"restaurant\", id))\n",
      "# plt.subplot(3,1,3)\n",
      "# plt.plot(qh_d[qh_d['util_service_point_id']==util_id][usage_columns].mean())\n",
      "# plt.yticks(np.linspace(0,max_val,5))\n",
      "# plt.xticks(range(0, 96, 16), times_of_day()[::16])\n",
      "\n",
      "# # single day/month graph\n",
      "# # id = 2000\n",
      "# # plt.plot(qh_d.iloc[id][usage_columns])\n",
      "# # plt.xticks(range(0, 96, 16), times_of_day()[::16])\n",
      "# # util_id = qh_d.iloc[id]['util_service_point_id']\n",
      "# # print util_id\n",
      "# # #plt.plot(qh_d.iloc[2][usage_columns])\n",
      "# # #plt.savefig('foo_%s_%d.png' % (\"restaurant\", id))\n",
      "# # plt.plot(qh_d[qh_d['util_service_point_id']==util_id][usage_columns].mean())"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 129
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# #single day/month graph\n",
      "# id = 0\n",
      "# #plt.plot(qh_d.iloc[id][usage_columns])\n",
      "# #plt.xticks(range(0, 96, 16), times_of_day()[::16])\n",
      "# util_id = 1137741705 #qh_d.iloc[id]['util_service_point_id']\n",
      "# print util_id\n",
      "# #plt.plot(qh_d.iloc[2][usage_columns])\n",
      "# #plt.savefig('foo_%s_%d.png' % (\"restaurant\", id))\n",
      "# mean_data = qh_d[qh_d['util_service_point_id']==util_id][usage_columns].mean()\n",
      "# plt.plot(mean_data)\n",
      "\n",
      "# switch_index = 1\n",
      "# switch_index_1 = 33\n",
      "# switch_index_2 = 69\n",
      "# plt.axvline(switch_index, color='red')\n",
      "# plt.axvline(switch_index_1, color='red')\n",
      "# plt.axvline(switch_index_2, color='red')\n",
      "# #plt.text(switch_index, day_values.min() + (day_values.max() - \n",
      "# #                                        day_values.min())*0.05, ' Closed', color='red',\n",
      "# #                                        verticalalignment='baseline', size='larger', weight='bold')"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 130
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# #single day/month graph\n",
      "# id = 0\n",
      "# #plt.plot(qh_d.iloc[id][usage_columns])\n",
      "# #plt.xticks(range(0, 96, 16), times_of_day()[::16])\n",
      "# util_id = 1137811605 #qh_d.iloc[id]['util_service_point_id']\n",
      "# print util_id\n",
      "# #plt.plot(qh_d.iloc[2][usage_columns])\n",
      "# #plt.savefig('foo_%s_%d.png' % (\"restaurant\", id))\n",
      "# mean_data = qh_d[qh_d['util_service_point_id']==util_id][usage_columns].mean()\n",
      "# plt.plot(mean_data)\n",
      "\n",
      "# switch_index = 1\n",
      "# switch_index_1 = 32\n",
      "# switch_index_2 = 69\n",
      "# plt.xticks(range(0, 96, 16), times_of_day()[::16])\n",
      "# plt.axvline(switch_index, color='red')\n",
      "# plt.axvline(switch_index_1, color='red')\n",
      "# plt.axvline(switch_index_2, color='red')\n",
      "# #plt.text(switch_index, day_values.min() + (day_values.max() - \n",
      "# #                                        day_values.min())*0.05, ' Closed', color='red',\n",
      "# #                                        verticalalignment='baseline', size='larger', weight='bold')"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 131
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# # state graph for restaurants \n",
      "# util_id = 8565778105 #qh_d.iloc[id]['util_service_point_id']\n",
      "# print util_id\n",
      "# mean_data = qh_r[qh_r['util_service_point_id']==util_id][usage_columns].mean()\n",
      "# plt.plot(mean_data)\n",
      "\n",
      "# switch_index = 1\n",
      "# switch_index_1 = 47\n",
      "# switch_index_3 = 89\n",
      "# plt.xticks(range(0, 96, 16), times_of_day()[::16])\n",
      "# plt.axvline(switch_index, color='red')\n",
      "# plt.axvline(switch_index_1, color='red')\n",
      "# plt.axvline(switch_index_2, color='red')\n",
      "# plt.axvline(switch_index_3, color='red')"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 132
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# #single day/month graph\n",
      "# id = 0\n",
      "# #plt.plot(qh_d.iloc[id][usage_columns])\n",
      "# #plt.xticks(range(0, 96, 16), times_of_day()[::16])\n",
      "# util_id = qh_d.iloc[id]['util_service_point_id']\n",
      "# print '1137741705'\n",
      "# #plt.plot(qh_d.iloc[2][usage_columns])\n",
      "# #plt.savefig('foo_%s_%d.png' % (\"restaurant\", id))\n",
      "# mean_data = qh_d[qh_d['util_service_point_id']==util_id][usage_columns].mean()\n",
      "# plt.plot(mean_data)\n",
      "\n",
      "# switch_index = 1\n",
      "# switch_index_1 = 33\n",
      "# switch_index_2 = 69\n",
      "# plt.axvline(switch_index, color='red')\n",
      "# plt.axvline(switch_index_1, color='red')\n",
      "# plt.axvline(switch_index_2, color='red')\n",
      "# #plt.text(switch_index, day_values.min() + (day_values.max() - \n",
      "# #                                        day_values.min())*0.05, ' Closed', color='red',\n",
      "# #                                        verticalalignment='baseline', size='larger', weight='bold')"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 133
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# Read in restaurant probabilities\n",
      "# restaurantP = pd.read_csv(\"data/restaurantProbability.csv\", sep=\",\")\n",
      "# july\n",
      "# restaurantP = pd.read_csv(\"data/JulRestaurantProbability.csv\", sep=\",\")\n",
      "# multiclass\n",
      "restaurantP = pd.read_csv(\"data/MulticlassRestaurantProb.csv\", sep=\",\")"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 134
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# Read in dentist probabilities\n",
      "# dentistP = pd.read_csv(\"data/dentistProbability.csv\", sep=\",\")\n",
      "# july\n",
      "# dentistP = pd.read_csv(\"data/JulDentistProbability.csv\", sep=\",\")\n",
      "# multiclass\n",
      "dentistP = pd.read_csv(\"data/MulticlassDentistProb.csv\", sep=\",\")"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 135
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# Read in school probabilities\n",
      "# multiclass\n",
      "# TODO schoolP = pd.read_csv(\"data/MulticlassSchoolProb.csv\", sep=\",\")"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 136
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# Read in restaurant labels\n",
      "# restaurantLabels = pd.read_csv(\"data/restaurantLabels.csv\", sep=\",\")\n",
      "# july\n",
      "# restaurantLabels = pd.read_csv(\"data/JulRestaurantLabels.csv\", sep=\",\")\n",
      "# multiclass\n",
      "restaurantLabels = pd.read_csv(\"data/MulticlassRestaurantLabels.csv\", sep=\",\")"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 137
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# Read in dentist labels\n",
      "# dentistLabels = pd.read_csv(\"data/dentistLabels.csv\", sep=\",\")\n",
      "# july\n",
      "# dentistLabels = pd.read_csv(\"data/JulDentistLabels.csv\", sep=\",\")\n",
      "# multiclass\n",
      "dentistLabels = pd.read_csv(\"data/MulticlassDentistLabels.csv\", sep=\",\")"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 138
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# Read in school labels\n",
      "schoolLabels = pd.read_csv(\"data/MulticlassSchoolLabels.csv\", sep=\",\")"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 139
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# Read in business types\n",
      "types = pd.read_csv(\"data/pge_smb_naics.tsv\", sep=\"\\t\")"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 140
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "three_class = False"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 141
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# Merging AMI and types datasets\n",
      "pre_merged_d = pd.merge(types, qh_d, on=\"util_service_point_id\")\n",
      "pre_merged_r = pd.merge(types, qh_r, on=\"util_service_point_id\")\n",
      "\n",
      "# schools\n",
      "if three_class is True:\n",
      "    pre_merged_s = pd.merge(types, qh_s, on=\"util_service_point_id\")"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 142
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# weekend cell\n",
      "pre_merged_d['weekday'] = pd.DatetimeIndex(pre_merged_d['date_value']).weekday\n",
      "pre_merged_r['weekday'] = pd.DatetimeIndex(pre_merged_r['date_value']).weekday\n",
      "\n",
      "# split out weekends\n",
      "pre_merged_d_weekday = pre_merged_d[pre_merged_d['weekday'] <= 4]\n",
      "pre_merged_d_weekend = pre_merged_d[pre_merged_d['weekday'] >= 5]\n",
      "\n",
      "pre_merged_r_weekday = pre_merged_r[pre_merged_r['weekday'] <= 4]\n",
      "pre_merged_r_weekend = pre_merged_r[pre_merged_r['weekday'] >= 5]\n",
      "\n",
      "# schools\n",
      "if three_class is True:\n",
      "    pre_merged_s['weekday'] = pd.DatetimeIndex(pre_merged_s['date_value']).weekday\n",
      "    pre_merged_s_weekday = pre_merged_s[pre_merged_s['weekday'] <= 4]\n",
      "    pre_merged_s_weekend = pre_merged_s[pre_merged_s['weekday'] >= 5]"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 143
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# Groupby a given business id to return one row per business\n",
      "merged_d_group = pre_merged_d.groupby(['util_service_point_id'], as_index=False)\n",
      "merged_r_group = pre_merged_r.groupby(['util_service_point_id'], as_index=False)\n",
      "\n",
      "# schools\n",
      "if three_class is True:\n",
      "    merged_s_group = pre_merged_s.groupby(['util_service_point_id'], as_index=False)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 144
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# weekend cell\n",
      "# Groupby splits for weekends and weekdays\n",
      "merged_d_group_day = pre_merged_d_weekday.groupby(['util_service_point_id'], as_index=False)\n",
      "merged_d_group_end = pre_merged_d_weekend.groupby(['util_service_point_id'], as_index=False)\n",
      "\n",
      "merged_r_group_day = pre_merged_r_weekday.groupby(['util_service_point_id'], as_index=False)\n",
      "merged_r_group_end = pre_merged_r_weekend.groupby(['util_service_point_id'], as_index=False)\n",
      "\n",
      "# schools\n",
      "if three_class is True:\n",
      "    merged_s_group_day = pre_merged_s_weekday.groupby(['util_service_point_id'], as_index=False)\n",
      "    merged_s_group_end = pre_merged_s_weekend.groupby(['util_service_point_id'], as_index=False)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 145
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "merged_d_series = merged_d_group.mean()\n",
      "merged_r_series = merged_r_group.mean()\n",
      "\n",
      "# schools\n",
      "if three_class is True:\n",
      "    merged_s_series = merged_s_group.mean()\n",
      "    \n",
      "# Loop controlling mean vs. single day\n",
      "# def get_series(average):\n",
      "#     if average == True:\n",
      "#         # Monthly average: take the average of all readings for a given business\n",
      "#         merged_d_series = merged_d_group.mean()\n",
      "#         merged_r_series = merged_r_group.mean()\n",
      "#     else:\n",
      "#         # Single day: use a single reading for a given business\n",
      "#         merged_d_series = merged_d_group.head(1)\n",
      "#         merged_r_series = merged_r_group.head(1)\n",
      "#     return merged_d_series, merged_r_series_day\n",
      "\n",
      "# # Output one row per business, either choose a single day or average over all days\n",
      "# average = True\n",
      "# (merged_d_series, merged_r_series) = get_series(average)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 146
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# weekend cell\n",
      "# take the average of weekend/day readings for a given business\n",
      "merged_d_series_day = merged_d_group_day.mean()\n",
      "merged_d_series_end = merged_d_group_end.mean()\n",
      "\n",
      "merged_r_series_day = merged_r_group_day.mean()\n",
      "merged_r_series_end = merged_r_group_end.mean()\n",
      "\n",
      "# schools\n",
      "if three_class is True:\n",
      "    merged_s_series_day = merged_s_group_day.mean()\n",
      "    merged_s_series_end = merged_s_group_end.mean()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 147
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# Converting series back into dataframe\n",
      "merged_d = pd.DataFrame(merged_d_series)\n",
      "merged_r = pd.DataFrame(merged_r_series)\n",
      "# schools\n",
      "if three_class is True:\n",
      "    merged_s = pd.DataFrame(merged_s_series)\n",
      "\n",
      "# Converting series back into dataframe, split for weekday/end\n",
      "merged_d_day = pd.DataFrame(merged_d_series_day)\n",
      "merged_d_end = pd.DataFrame(merged_d_series_end)\n",
      "\n",
      "merged_r_day = pd.DataFrame(merged_r_series_day)\n",
      "merged_r_end = pd.DataFrame(merged_r_series_end)\n",
      "\n",
      "# schools\n",
      "if three_class is True:\n",
      "    merged_s_day = pd.DataFrame(merged_s_series_day)\n",
      "    merged_s_end = pd.DataFrame(merged_s_series_end)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 148
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# Add name4 column\n",
      "merged_d[\"name4\"] = \"Offices of Dentists\"\n",
      "merged_r[\"name4\"] = \"Food Services and Drinking Places\"\n",
      "\n",
      "# schools\n",
      "if three_class is True:\n",
      "    merged_s[\"name4\"] = \"Elementary and Secondary Schools\""
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 149
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# Get all dentists and restaurants\n",
      "dentists = merged_d[merged_d[\"name4\"]==\"Offices of Dentists\"]\n",
      "restaurants = merged_r[merged_r[\"name4\"]==\"Food Services and Drinking Places\"]\n",
      "dentists = merged_d.dropna()\n",
      "restaurants = merged_r.dropna() # TODO shouldn't this be x = x.dropna()\n",
      "\n",
      "# schools\n",
      "if three_class is True:\n",
      "    schools = merged_s[merged_s[\"name4\"]==\"Elementary and Secondary Schools\"]\n",
      "    schools = merged_s.dropna()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 150
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# weekend cell\n",
      "# get all dentists and restaurants\n",
      "merged_d_day[\"name4\"] = \"Offices of Dentists\"\n",
      "merged_d_end[\"name4\"] = \"Offices of Dentists\"\n",
      "merged_r_day[\"name4\"] = \"Food Services and Drinking Places\"\n",
      "merged_r_end[\"name4\"] = \"Food Services and Drinking Places\"\n",
      "\n",
      "dentists_day = merged_d_day.dropna()\n",
      "dentists_end = merged_d_end.dropna()\n",
      "restaurants_day = merged_r_day.dropna()\n",
      "restaurants_end = merged_r_end.dropna()\n",
      "\n",
      "# schools\n",
      "if three_class is True:\n",
      "    merged_s_day[\"name4\"] = \"Elementary and Secondary Schools\"\n",
      "    merged_s_end[\"name4\"] = \"Elementary and Secondary Schools\"\n",
      "    schools_day = merged_s_day.dropna()\n",
      "    schools_end = merged_s_end.dropna()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 151
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# weekend cell\n",
      "# weekday merge into giant feature table\n",
      "weekend_and_weekday_d = pd.merge(dentists_day, dentists_end, suffixes=[\"_day\", \"_end\"], on=\"util_service_point_id\")\n",
      "weekend_and_weekday_r = pd.merge(restaurants_day, restaurants_end, suffixes=[\"_day\", \"_end\"], on=\"util_service_point_id\")\n",
      "\n",
      "# schools\n",
      "if three_class is True:\n",
      "    weekend_and_weekday_s = pd.merge(schools_day, schools_end, suffixes=[\"_day\", \"_end\"], on=\"util_service_point_id\")\n",
      "    \n"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 152
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# Merge probability column in, restaurants\n",
      "merged_with_restaurants = restaurants.merge(restaurantP, on='util_service_point_id', how=\"inner\")\n",
      "merged_with_restaurants = merged_with_restaurants.merge(restaurantLabels, on='util_service_point_id', how=\"inner\")\n",
      "merged_with_restaurants = merged_with_restaurants.merge(weekend_and_weekday_r, on='util_service_point_id', how=\"inner\")"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 153
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# Merge probability column in, dentists\n",
      "merged_with_dentists = dentists.merge(dentistP, on='util_service_point_id', how=\"inner\")\n",
      "merged_with_dentists = merged_with_dentists.merge(dentistLabels, on='util_service_point_id', how=\"inner\")\n",
      "merged_with_dentists = merged_with_dentists.merge(weekend_and_weekday_d, on='util_service_point_id', how='inner')\n"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 154
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# schools\n",
      "if three_class is True:\n",
      "    # Merge probability column in, schools\n",
      "    merged_with_schools = schools.merge(schoolP, on='util_service_point_id', how=\"inner\")\n",
      "    merged_with_schools = merged_with_schools.merge(schoolLabels, on='util_service_point_id', how=\"inner\")\n",
      "    merged_with_schools = merged_with_schools.merge(weekend_and_weekday_s, on='util_service_point_id', how='inner')"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 155
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# Concat dentists and restaurants\n",
      "min_len = min(len(merged_with_dentists), len(merged_with_restaurants))\n",
      "smb = pd.concat([merged_with_dentists[:min_len], merged_with_restaurants[:min_len]])\n",
      "smb.index = range(len(smb))\n",
      "# schools\n",
      "if three_class is True:\n",
      "    min_len = min(len(smb), len(merged_with_schools))\n",
      "    smb = pd.concat([merged_with_schools[:min_len], smb[:min_len]])\n",
      "    smb.index = range(len(smb))"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 156
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# Get all features\n",
      "smb.fillna(0.0, inplace=True)\n",
      "\n",
      "# Two Class\n",
      "# features = smb[list(smb.columns[19:115]) + ['dentist', 'probabilityDentist', 'name4']\n",
      "#                + list(smb.columns[137:233]) + list(smb.columns[253:349])]\n",
      "\n",
      "# Three Class\n",
      "features = smb[list(smb.columns[19:115]) + ['label', 'probabilityDentist', \n",
      "                                            'probabilitySchool', 'name4']\n",
      "                + list(smb.columns[138:234]) + list(smb.columns[254:350])]"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 157
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# monthly\n",
      "# features_d = dentists.iloc[:,19:115]\n",
      "# features_r = restaurants.iloc[:,19:115]\n",
      "\n",
      "# single day\n",
      "# features_d = dentists.iloc[:,25:122]\n",
      "# features_r = restaurants.iloc[:,25:122]"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 158
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# Shuffle features\n",
      "np.random.seed(0)\n",
      "features = features.reindex(np.random.permutation(features.index)) "
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 159
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# Define usage column feature sets\n",
      "overall_usage_columns = [u'usage_value_000', u'usage_value_015', u'usage_value_030', u'usage_value_045', u'usage_value_100', u'usage_value_115', u'usage_value_130', u'usage_value_145', u'usage_value_200', u'usage_value_215', u'usage_value_230', u'usage_value_245', u'usage_value_300', u'usage_value_315', u'usage_value_330', u'usage_value_345', u'usage_value_400', u'usage_value_415', u'usage_value_430', u'usage_value_445', u'usage_value_500', u'usage_value_515', u'usage_value_530', u'usage_value_545', u'usage_value_600', u'usage_value_615', u'usage_value_630', u'usage_value_645', u'usage_value_700', u'usage_value_715', u'usage_value_730', u'usage_value_745', u'usage_value_800', u'usage_value_815', u'usage_value_830', u'usage_value_845', u'usage_value_900', u'usage_value_915', u'usage_value_930', u'usage_value_945', u'usage_value_1000', u'usage_value_1015', u'usage_value_1030', u'usage_value_1045', u'usage_value_1100', u'usage_value_1115', u'usage_value_1130', u'usage_value_1145', u'usage_value_1200', u'usage_value_1215', u'usage_value_1230', u'usage_value_1245', u'usage_value_1300', u'usage_value_1315', u'usage_value_1330', u'usage_value_1345', u'usage_value_1400', u'usage_value_1415', u'usage_value_1430', u'usage_value_1445', u'usage_value_1500', u'usage_value_1515', u'usage_value_1530', u'usage_value_1545', u'usage_value_1600', u'usage_value_1615', u'usage_value_1630', u'usage_value_1645', u'usage_value_1700', u'usage_value_1715', u'usage_value_1730', u'usage_value_1745', u'usage_value_1800', u'usage_value_1815', u'usage_value_1830', u'usage_value_1845', u'usage_value_1900', u'usage_value_1915', u'usage_value_1930', u'usage_value_1945', u'usage_value_2000', u'usage_value_2015', u'usage_value_2030', u'usage_value_2045', u'usage_value_2100', u'usage_value_2115', u'usage_value_2130', u'usage_value_2145', u'usage_value_2200', u'usage_value_2215', u'usage_value_2230', u'usage_value_2245', u'usage_value_2300', u'usage_value_2315', u'usage_value_2330', u'usage_value_2345']\n",
      "weekday_usage_columns = [u'usage_value_000_day', u'usage_value_015_day', u'usage_value_030_day', u'usage_value_045_day', u'usage_value_100_day', u'usage_value_115_day', u'usage_value_130_day', u'usage_value_145_day', u'usage_value_200_day', u'usage_value_215_day', u'usage_value_230_day', u'usage_value_245_day', u'usage_value_300_day', u'usage_value_315_day', u'usage_value_330_day', u'usage_value_345_day', u'usage_value_400_day', u'usage_value_415_day', u'usage_value_430_day', u'usage_value_445_day', u'usage_value_500_day', u'usage_value_515_day', u'usage_value_530_day', u'usage_value_545_day', u'usage_value_600_day', u'usage_value_615_day', u'usage_value_630_day', u'usage_value_645_day', u'usage_value_700_day', u'usage_value_715_day', u'usage_value_730_day', u'usage_value_745_day', u'usage_value_800_day', u'usage_value_815_day', u'usage_value_830_day', u'usage_value_845_day', u'usage_value_900_day', u'usage_value_915_day', u'usage_value_930_day', u'usage_value_945_day', u'usage_value_1000_day', u'usage_value_1015_day', u'usage_value_1030_day', u'usage_value_1045_day', u'usage_value_1100_day', u'usage_value_1115_day', u'usage_value_1130_day', u'usage_value_1145_day', u'usage_value_1200_day', u'usage_value_1215_day', u'usage_value_1230_day', u'usage_value_1245_day', u'usage_value_1300_day', u'usage_value_1315_day', u'usage_value_1330_day', u'usage_value_1345_day', u'usage_value_1400_day', u'usage_value_1415_day', u'usage_value_1430_day', u'usage_value_1445_day', u'usage_value_1500_day', u'usage_value_1515_day', u'usage_value_1530_day', u'usage_value_1545_day', u'usage_value_1600_day', u'usage_value_1615_day', u'usage_value_1630_day', u'usage_value_1645_day', u'usage_value_1700_day', u'usage_value_1715_day', u'usage_value_1730_day', u'usage_value_1745_day', u'usage_value_1800_day', u'usage_value_1815_day', u'usage_value_1830_day', u'usage_value_1845_day', u'usage_value_1900_day', u'usage_value_1915_day', u'usage_value_1930_day', u'usage_value_1945_day', u'usage_value_2000_day', u'usage_value_2015_day', u'usage_value_2030_day', u'usage_value_2045_day', u'usage_value_2100_day', u'usage_value_2115_day', u'usage_value_2130_day', u'usage_value_2145_day', u'usage_value_2200_day', u'usage_value_2215_day', u'usage_value_2230_day', u'usage_value_2245_day', u'usage_value_2300_day', u'usage_value_2315_day', u'usage_value_2330_day', u'usage_value_2345_day']\n",
      "weekend_usage_columns = [u'usage_value_000_end', u'usage_value_015_end', u'usage_value_030_end', u'usage_value_045_end', u'usage_value_100_end', u'usage_value_115_end', u'usage_value_130_end', u'usage_value_145_end', u'usage_value_200_end', u'usage_value_215_end', u'usage_value_230_end', u'usage_value_245_end', u'usage_value_300_end', u'usage_value_315_end', u'usage_value_330_end', u'usage_value_345_end', u'usage_value_400_end', u'usage_value_415_end', u'usage_value_430_end', u'usage_value_445_end', u'usage_value_500_end', u'usage_value_515_end', u'usage_value_530_end', u'usage_value_545_end', u'usage_value_600_end', u'usage_value_615_end', u'usage_value_630_end', u'usage_value_645_end', u'usage_value_700_end', u'usage_value_715_end', u'usage_value_730_end', u'usage_value_745_end', u'usage_value_800_end', u'usage_value_815_end', u'usage_value_830_end', u'usage_value_845_end', u'usage_value_900_end', u'usage_value_915_end', u'usage_value_930_end', u'usage_value_945_end', u'usage_value_1000_end', u'usage_value_1015_end', u'usage_value_1030_end', u'usage_value_1045_end', u'usage_value_1100_end', u'usage_value_1115_end', u'usage_value_1130_end', u'usage_value_1145_end', u'usage_value_1200_end', u'usage_value_1215_end', u'usage_value_1230_end', u'usage_value_1245_end', u'usage_value_1300_end', u'usage_value_1315_end', u'usage_value_1330_end', u'usage_value_1345_end', u'usage_value_1400_end', u'usage_value_1415_end', u'usage_value_1430_end', u'usage_value_1445_end', u'usage_value_1500_end', u'usage_value_1515_end', u'usage_value_1530_end', u'usage_value_1545_end', u'usage_value_1600_end', u'usage_value_1615_end', u'usage_value_1630_end', u'usage_value_1645_end', u'usage_value_1700_end', u'usage_value_1715_end', u'usage_value_1730_end', u'usage_value_1745_end', u'usage_value_1800_end', u'usage_value_1815_end', u'usage_value_1830_end', u'usage_value_1845_end', u'usage_value_1900_end', u'usage_value_1915_end', u'usage_value_1930_end', u'usage_value_1945_end', u'usage_value_2000_end', u'usage_value_2015_end', u'usage_value_2030_end', u'usage_value_2045_end', u'usage_value_2100_end', u'usage_value_2115_end', u'usage_value_2130_end', u'usage_value_2145_end', u'usage_value_2200_end', u'usage_value_2215_end', u'usage_value_2230_end', u'usage_value_2245_end', u'usage_value_2300_end', u'usage_value_2315_end', u'usage_value_2330_end', u'usage_value_2345_end']"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 160
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# Define regression steps\n",
      "def run_regression(features, seed=0, n_folds=10, standardize_feats=False):\n",
      "    num_correct = 0\n",
      "    for train_index, test_index in cross_validation.KFold(n=len(features), n_folds=n_folds, shuffle=True, random_state=seed):\n",
      "        # features\n",
      "        train_features = features.iloc[list(train_index)]\n",
      "        test_features = features.iloc[list(test_index)]\n",
      "\n",
      "        # labels\n",
      "        train_labels = train_features['name4']\n",
      "        test_labels = test_features['name4']\n",
      "\n",
      "        # drop labels from feature set\n",
      "        train_features.drop('name4', axis=1, inplace=True)\n",
      "        test_features.drop('name4', axis=1, inplace=True)\n",
      "\n",
      "        # standardize\n",
      "        if standardize_feats:\n",
      "            train_mean = train_features.mean()\n",
      "            train_mean_zero = train_features - train_mean\n",
      "            train_std = train_mean_zero.std()\n",
      "            train_std_one = train_mean_zero / train_std\n",
      "            train_features = train_std_one\n",
      "            \n",
      "            test_mean_zero = test_features - train_mean\n",
      "            test_std_one = test_mean_zero / train_std\n",
      "            test_features = test_std_one\n",
      "        \n",
      "        # fit \n",
      "        lr = linear_model.LogisticRegression()\n",
      "        lr.fit(train_features, train_labels)\n",
      "\n",
      "        # predict\n",
      "        predictions = lr.predict(test_features)\n",
      "\n",
      "        # score\n",
      "        correct = predictions==test_labels\n",
      "        num_correct += sum(correct)\n",
      "    \n",
      "    accuracy = num_correct/float(len(features))\n",
      "    print accuracy\n",
      "    return accuracy"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 161
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# Define normalization steps\n",
      "def do_normalization(normalize, features):\n",
      "    usage_features = features[overall_usage_columns]\n",
      "    sum_column = usage_features.sum(axis=1)\n",
      "    usage_features = usage_features.div(sum_column,axis='index')\n",
      "    usage_features.fillna(0.0,inplace=True)\n",
      "    if normalize:\n",
      "        features[overall_usage_columns] = usage_features\n",
      "    features['usage_sum'] = sum_column\n",
      "    return features"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 162
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# Define features sets \n",
      "def pick_features(features, feature_set):\n",
      "    if feature_set == 'just_usage':\n",
      "        return features[overall_usage_columns + ['name4']]\n",
      "    elif feature_set == 'usage_sum':\n",
      "        return features[overall_usage_columns + ['usage_sum', 'name4']]\n",
      "    elif feature_set == 'just_prob':\n",
      "        # features for multiclass\n",
      "        # return features[['probabilityDentist', 'name4']]\n",
      "        return features[['probabilityDentist', 'probabilitySchool', 'name4']]\n",
      "    elif feature_set == 'usage_prob':\n",
      "        # features for multiclass\n",
      "        # return features[overall_usage_columns + ['probabilityDentist', 'name4']]\n",
      "        return features[overall_usage_columns + ['probabilityDentist', 'probabilitySchool', 'name4']]\n",
      "    elif feature_set == 'usage_label':\n",
      "        # features for multiclass\n",
      "        # return features[overall_usage_columns + ['dentist', 'name4']]\n",
      "        return features[overall_usage_columns + ['label', 'name4']]\n",
      "    elif feature_set == 'usage_prob_label':\n",
      "        # features for multiclass\n",
      "        # return features[overall_usage_columns + ['probabilityDentist', 'dentist', 'name4']]\n",
      "        return features[overall_usage_columns + ['probabilityDentist', 'probabilitySchool', 'label', 'name4']]\n",
      "    elif feature_set == 'just_usage_w':\n",
      "        return features[weekday_usage_columns + weekend_usage_columns + ['name4']]\n",
      "    elif feature_set == 'usage_w_prob':\n",
      "        # features for multiclass\n",
      "        # return features[weekday_usage_columns + weekend_usage_columns + ['probabilityDentist', 'name4']]\n",
      "        return features[weekday_usage_columns + weekend_usage_columns + ['probabilityDentist','probabilitySchool', 'name4']]\n",
      "    elif feature_set == 'usage_w_label':\n",
      "        # features for multiclass\n",
      "        # return features[weekday_usage_columns + weekend_usage_columns + ['dentist', 'name4']]\n",
      "        return features[weekday_usage_columns + weekend_usage_columns + ['label', 'name4']]\n",
      "    elif feature_set == 'usage_w_prob_label':\n",
      "        # features for multiclass\n",
      "        # return features[weekday_usage_columns + weekend_usage_columns + ['probabilityDentist', 'dentist', 'name4']]\n",
      "        return features[weekday_usage_columns + weekend_usage_columns + ['probabilityDentist','probabilitySchool', 'label', 'name4']]"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 163
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "feature_sets = ['just_usage', 'usage_sum', 'just_prob', 'usage_prob', 'usage_label', 'usage_prob_label', \n",
      "                'just_usage_w', 'usage_w_prob', 'usage_w_label', 'usage_w_prob_label']\n",
      "\n",
      "normalize_usage_values = [True, False]\n",
      "standardize_values = [True, False]\n",
      "\n",
      "# Run experiments\n",
      "accuracies = []\n",
      "names = []\n",
      "seed = 0\n",
      "n_folds = 10\n",
      "for standardize in standardize_values:\n",
      "    for normalize_usage in normalize_usage_values:\n",
      "        experiment_features_normalization = features.copy()\n",
      "        experiment_features_normalization = do_normalization(normalize_usage, experiment_features_normalization)\n",
      "        for feature_set in feature_sets:\n",
      "            experiment_features = experiment_features_normalization.copy()\n",
      "            experiment_features = pick_features(experiment_features, feature_set)\n",
      "            \n",
      "            # print accuracies\n",
      "            print \"Feature set: %s\" % feature_set\n",
      "            print \"Standardize =\", standardize\n",
      "            print \"Normalization =\",normalize_usage\n",
      "            names.append(\"%s_%s_%s\" % (feature_set, normalize, standardize))\n",
      "            accuracies.append(run_regression(experiment_features, seed, n_folds, standardize))\n",
      "            print '\\n'\n",
      "\n",
      "# plot accuracies\n",
      "plt.bar(range(len(accuracies)), accuracies)\n",
      "plt.title(\"Accuracy Histogram\")\n",
      "plt.xlabel(\"Feature Set\")\n",
      "plt.ylabel(\"Accuracy\")\n",
      "plt.xticks(range(len(accuracies)), names, rotation='vertical') \n",
      "plt.ylim([.75,1])\n",
      "plt.show()\n",
      "\n",
      "# print interesting accuracies\n",
      "# print len(accuracies)\n",
      "# print len(set(accuracies))\n",
      "# interesting_accuracies = list(set(accuracies))\n",
      "# plt.bar(range(len(interesting_accuracies)), interesting_accuracies)\n",
      "# plt.title(\"Accuracy Histogram\")\n",
      "# plt.xlabel(\"Feature Set\")\n",
      "# plt.ylabel(\"Accuracy\")\n",
      "# plt.xticks(range(len(interesting_accuracies)), names, rotation='vertical') \n",
      "# plt.ylim([.75,1])\n",
      "# plt.show()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "Feature set: just_usage\n",
        "Standardize = True\n",
        "Normalization = True\n",
        "0.858252427184"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_sum\n",
        "Standardize = True\n",
        "Normalization = True\n",
        "0.877669902913"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: just_prob\n",
        "Standardize = True\n",
        "Normalization = True\n",
        "0.757281553398"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_prob\n",
        "Standardize = True\n",
        "Normalization = True\n",
        "0.86213592233"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_label\n",
        "Standardize = True\n",
        "Normalization = True\n",
        "0.863106796117"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_prob_label\n",
        "Standardize = True\n",
        "Normalization = True\n",
        "0.863106796117"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: just_usage_w\n",
        "Standardize = True\n",
        "Normalization = True\n",
        "0.890291262136"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_w_prob\n",
        "Standardize = True\n",
        "Normalization = True\n",
        "0.888349514563"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_w_label\n",
        "Standardize = True\n",
        "Normalization = True\n",
        "0.885436893204"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_w_prob_label\n",
        "Standardize = True\n",
        "Normalization = True\n",
        "0.888349514563"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: just_usage\n",
        "Standardize = True\n",
        "Normalization = False\n",
        "0.86213592233"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_sum\n",
        "Standardize = True\n",
        "Normalization = False\n",
        "0.86213592233"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: just_prob\n",
        "Standardize = True\n",
        "Normalization = False\n",
        "0.757281553398"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_prob\n",
        "Standardize = True\n",
        "Normalization = False\n",
        "0.86213592233"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_label\n",
        "Standardize = True\n",
        "Normalization = False\n",
        "0.861165048544"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_prob_label\n",
        "Standardize = True\n",
        "Normalization = False\n",
        "0.86213592233"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: just_usage_w\n",
        "Standardize = True\n",
        "Normalization = False\n",
        "0.890291262136"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_w_prob\n",
        "Standardize = True\n",
        "Normalization = False\n",
        "0.888349514563"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_w_label\n",
        "Standardize = True\n",
        "Normalization = False\n",
        "0.885436893204"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_w_prob_label\n",
        "Standardize = True\n",
        "Normalization = False\n",
        "0.888349514563"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: just_usage\n",
        "Standardize = False\n",
        "Normalization = True\n",
        "0.855339805825"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_sum\n",
        "Standardize = False\n",
        "Normalization = True\n",
        "0.799029126214"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: just_prob\n",
        "Standardize = False\n",
        "Normalization = True\n",
        "0.757281553398"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_prob\n",
        "Standardize = False\n",
        "Normalization = True\n",
        "0.757281553398"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_label\n",
        "Standardize = False\n",
        "Normalization = True\n",
        "0.797087378641"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_prob_label\n",
        "Standardize = False\n",
        "Normalization = True\n",
        "0.757281553398"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: just_usage_w\n",
        "Standardize = False\n",
        "Normalization = True\n",
        "0.888349514563"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_w_prob\n",
        "Standardize = False\n",
        "Normalization = True\n",
        "0.887378640777"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_w_label\n",
        "Standardize = False\n",
        "Normalization = True\n",
        "0.88640776699"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_w_prob_label\n",
        "Standardize = False\n",
        "Normalization = True\n",
        "0.88932038835"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: just_usage\n",
        "Standardize = False\n",
        "Normalization = False\n",
        "0.864077669903"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_sum\n",
        "Standardize = False\n",
        "Normalization = False\n",
        "0.863106796117"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: just_prob\n",
        "Standardize = False\n",
        "Normalization = False\n",
        "0.757281553398"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_prob\n",
        "Standardize = False\n",
        "Normalization = False\n",
        "0.856310679612"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_label\n",
        "Standardize = False\n",
        "Normalization = False\n",
        "0.865048543689"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_prob_label\n",
        "Standardize = False\n",
        "Normalization = False\n",
        "0.859223300971"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: just_usage_w\n",
        "Standardize = False\n",
        "Normalization = False\n",
        "0.888349514563"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_w_prob\n",
        "Standardize = False\n",
        "Normalization = False\n",
        "0.887378640777"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_w_label\n",
        "Standardize = False\n",
        "Normalization = False\n",
        "0.88640776699"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n",
        "Feature set: usage_w_prob_label\n",
        "Standardize = False\n",
        "Normalization = False\n",
        "0.88932038835"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "\n",
        "\n"
       ]
      },
      {
       "metadata": {},
       "output_type": "display_data",
       "png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAJWCAYAAAC3YZtUAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
       "text": [
        "<matplotlib.figure.Figure at 0x1124e7610>"
       ]
      }
     ],
     "prompt_number": 164
    }
   ],
   "metadata": {}
  }
 ]
}