GoodPrediction.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import math\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import os\n",
    "import tempfile\n",
    "\n",
    "from scipy.spatial.distance import pdist, squareform\n",
    "import scipy.io as sio\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "trainPredict = np.load('trainPredict.npy')\n",
    "train_result = np.sum(trainPredict, axis=1)\n",
    "testPredict = np.load('testPredict.npy')\n",
    "test_result = np.sum(testPredict, axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##  Get distributions of Dest by Source,Time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "orderdf = pd.read_csv('ignored_assets/orders.csv', parse_dates = [6])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Create map of driver IDs\n",
    "driver_ids=orderdf['driver_id'].unique()\n",
    "\n",
    "# create map of core stations\n",
    "core_stations = orderdf['start_district_hash'].unique()\n",
    "core_stations.sort()\n",
    "\n",
    "#Index stuff\n",
    "orderdf_f=orderdf[(~pd.isnull(orderdf['driver_id'])) & orderdf['dest_district_hash'].isin(core_stations)]\n",
    "myindex=pd.MultiIndex.from_arrays([orderdf_f['driver_id'],orderdf_f['timestamp']],names=('driver_id','timestamp'))\n",
    "\n",
    "orderdf2=orderdf_f.set_index(myindex)\n",
    "orderdf_s=orderdf2.sort_index()\n",
    "\n",
    "orderdf_s['time_bucket'] = orderdf_s['timestamp'].dt.round('5min')\n",
    "orderdf_s['pax_arrival_time'] = orderdf_s['timestamp'] + orderdf_s['expected_travel_time'] * pd.Timedelta('1 minute')\n",
    "orderdf_s['time_bucket_arrival'] = orderdf_s['pax_arrival_time'].dt.round('5min')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "stepsize = pd.Timedelta('5 min')\n",
    "horizon = pd.Timedelta('2 hour')\n",
    "start = pd.to_datetime('2016-01-20 00:00:00')\n",
    "end = pd.to_datetime('2016-01-21 00:00:00')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "mask = (orderdf_s['time_bucket'] >= start) & (orderdf_s['time_bucket'] < end)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "dates = [start + j*stepsize for j in range(int(horizon / stepsize))]\n",
    "iterables = [dates, core_stations]\n",
    "mindex = pd.MultiIndex.from_product(iterables, names=['time_bucket', 'start_district_hash'])\n",
    "mcolumns = pd.MultiIndex.from_arrays([core_stations], names=['dest_district_hash'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "base = pd.DataFrame(index=mindex, columns=mcolumns).fillna(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "pt = (base + pd.pivot_table(orderdf_s[mask], index=['time_bucket','start_district_hash'], \n",
    "               columns=['dest_district_hash'], aggfunc=len, \n",
    "               values='order_id', fill_value=0)).fillna(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "prior = pd.pivot_table(orderdf_s, index=['start_district_hash'], \n",
    "               columns=['dest_district_hash'], aggfunc=len, \n",
    "               values='order_id', fill_value=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#weight of the prior (based on hours and days of samples)\n",
    "ndays = 20.\n",
    "nhours = 24.\n",
    "w = 1 / (ndays * nhours)\n",
    "\n",
    "# sum both\n",
    "count = pt + w*prior\n",
    "\n",
    "# get posterior\n",
    "posterior = count.div( count.sum(axis=1), axis = 0).fillna(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>dest_district_hash</th>\n",
       "      <th>08232402614a9b48895cc3d0aeb0e9f2</th>\n",
       "      <th>08f5b445ec6b29deba62e6fd8b0325a6</th>\n",
       "      <th>0a5fef95db34383403d11cb6af937309</th>\n",
       "      <th>1afd7afbc81ecc1b13886a569d869e8a</th>\n",
       "      <th>1c60154546102e6525f68cb4f31e0657</th>\n",
       "      <th>1cbfbdd079ef93e74405c53fcfff8567</th>\n",
       "      <th>1ecbb52d73c522f184a6fc53128b1ea1</th>\n",
       "      <th>2301bc920194c95cf0c7486e5675243c</th>\n",
       "      <th>2350be163432e42270d2670cb3c02f80</th>\n",
       "      <th>2407d482f0ffa22a947068f2551fe62c</th>\n",
       "      <th>...</th>\n",
       "      <th>d4ec2125aff74eded207d2d915ef682f</th>\n",
       "      <th>d524868ce69cb9db10fc5af177fb9423</th>\n",
       "      <th>d5cb17978de290c56e84c9cf97e63186</th>\n",
       "      <th>dd8d3b9665536d6e05b29c2648c0e69a</th>\n",
       "      <th>de092beab9305613aca8f79d7d7224e7</th>\n",
       "      <th>f2c8c4bb99e6377d21de71275afd6cd2</th>\n",
       "      <th>f47f35242ed40655814bc086d7514046</th>\n",
       "      <th>f9280c5dab6910ed44e518248048b9fe</th>\n",
       "      <th>fc34648599753c9e74ab238e9a4a07ad</th>\n",
       "      <th>fff4e8465d1e12621bc361276b6217cf</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>time_bucket</th>\n",
       "      <th>start_district_hash</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">2016-01-20</th>\n",
       "      <th>08232402614a9b48895cc3d0aeb0e9f2</th>\n",
       "      <td>0.123467</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.001635</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000818</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.004906</td>\n",
       "      <td>0.013900</td>\n",
       "      <td>0.001635</td>\n",
       "      <td>...</td>\n",
       "      <td>0.048651</td>\n",
       "      <td>0.025348</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000409</td>\n",
       "      <td>0.035568</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.002862</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.012265</td>\n",
       "      <td>0.001226</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>08f5b445ec6b29deba62e6fd8b0325a6</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.042336</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.090235</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000618</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.052225</td>\n",
       "      <td>...</td>\n",
       "      <td>0.026885</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.007417</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000927</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0a5fef95db34383403d11cb6af937309</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.548872</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.007519</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.015038</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.090226</td>\n",
       "      <td>0.007519</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1afd7afbc81ecc1b13886a569d869e8a</th>\n",
       "      <td>0.000038</td>\n",
       "      <td>0.001509</td>\n",
       "      <td>0.000023</td>\n",
       "      <td>0.332601</td>\n",
       "      <td>0.000034</td>\n",
       "      <td>0.000625</td>\n",
       "      <td>0.000015</td>\n",
       "      <td>0.000079</td>\n",
       "      <td>0.000132</td>\n",
       "      <td>0.207689</td>\n",
       "      <td>...</td>\n",
       "      <td>0.070296</td>\n",
       "      <td>0.000015</td>\n",
       "      <td>0.000151</td>\n",
       "      <td>0.001942</td>\n",
       "      <td>0.000004</td>\n",
       "      <td>0.000026</td>\n",
       "      <td>0.000241</td>\n",
       "      <td>0.006696</td>\n",
       "      <td>0.000094</td>\n",
       "      <td>0.000004</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1c60154546102e6525f68cb4f31e0657</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000087</td>\n",
       "      <td>0.000346</td>\n",
       "      <td>0.022684</td>\n",
       "      <td>0.000087</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000693</td>\n",
       "      <td>0.000433</td>\n",
       "      <td>0.000519</td>\n",
       "      <td>...</td>\n",
       "      <td>0.007879</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.001645</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000087</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000952</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 66 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "dest_district_hash                            08232402614a9b48895cc3d0aeb0e9f2  \\\n",
       "time_bucket start_district_hash                                                  \n",
       "2016-01-20  08232402614a9b48895cc3d0aeb0e9f2                          0.123467   \n",
       "            08f5b445ec6b29deba62e6fd8b0325a6                          0.000000   \n",
       "            0a5fef95db34383403d11cb6af937309                          0.000000   \n",
       "            1afd7afbc81ecc1b13886a569d869e8a                          0.000038   \n",
       "            1c60154546102e6525f68cb4f31e0657                          0.000000   \n",
       "\n",
       "dest_district_hash                            08f5b445ec6b29deba62e6fd8b0325a6  \\\n",
       "time_bucket start_district_hash                                                  \n",
       "2016-01-20  08232402614a9b48895cc3d0aeb0e9f2                          0.000000   \n",
       "            08f5b445ec6b29deba62e6fd8b0325a6                          0.042336   \n",
       "            0a5fef95db34383403d11cb6af937309                          0.000000   \n",
       "            1afd7afbc81ecc1b13886a569d869e8a                          0.001509   \n",
       "            1c60154546102e6525f68cb4f31e0657                          0.000000   \n",
       "\n",
       "dest_district_hash                            0a5fef95db34383403d11cb6af937309  \\\n",
       "time_bucket start_district_hash                                                  \n",
       "2016-01-20  08232402614a9b48895cc3d0aeb0e9f2                          0.000000   \n",
       "            08f5b445ec6b29deba62e6fd8b0325a6                          0.000000   \n",
       "            0a5fef95db34383403d11cb6af937309                          0.548872   \n",
       "            1afd7afbc81ecc1b13886a569d869e8a                          0.000023   \n",
       "            1c60154546102e6525f68cb4f31e0657                          0.000087   \n",
       "\n",
       "dest_district_hash                            1afd7afbc81ecc1b13886a569d869e8a  \\\n",
       "time_bucket start_district_hash                                                  \n",
       "2016-01-20  08232402614a9b48895cc3d0aeb0e9f2                          0.001635   \n",
       "            08f5b445ec6b29deba62e6fd8b0325a6                          0.090235   \n",
       "            0a5fef95db34383403d11cb6af937309                          0.000000   \n",
       "            1afd7afbc81ecc1b13886a569d869e8a                          0.332601   \n",
       "            1c60154546102e6525f68cb4f31e0657                          0.000346   \n",
       "\n",
       "dest_district_hash                            1c60154546102e6525f68cb4f31e0657  \\\n",
       "time_bucket start_district_hash                                                  \n",
       "2016-01-20  08232402614a9b48895cc3d0aeb0e9f2                          0.000000   \n",
       "            08f5b445ec6b29deba62e6fd8b0325a6                          0.000000   \n",
       "            0a5fef95db34383403d11cb6af937309                          0.000000   \n",
       "            1afd7afbc81ecc1b13886a569d869e8a                          0.000034   \n",
       "            1c60154546102e6525f68cb4f31e0657                          0.022684   \n",
       "\n",
       "dest_district_hash                            1cbfbdd079ef93e74405c53fcfff8567  \\\n",
       "time_bucket start_district_hash                                                  \n",
       "2016-01-20  08232402614a9b48895cc3d0aeb0e9f2                          0.000818   \n",
       "            08f5b445ec6b29deba62e6fd8b0325a6                          0.000618   \n",
       "            0a5fef95db34383403d11cb6af937309                          0.007519   \n",
       "            1afd7afbc81ecc1b13886a569d869e8a                          0.000625   \n",
       "            1c60154546102e6525f68cb4f31e0657                          0.000087   \n",
       "\n",
       "dest_district_hash                            1ecbb52d73c522f184a6fc53128b1ea1  \\\n",
       "time_bucket start_district_hash                                                  \n",
       "2016-01-20  08232402614a9b48895cc3d0aeb0e9f2                          0.000000   \n",
       "            08f5b445ec6b29deba62e6fd8b0325a6                          0.000000   \n",
       "            0a5fef95db34383403d11cb6af937309                          0.000000   \n",
       "            1afd7afbc81ecc1b13886a569d869e8a                          0.000015   \n",
       "            1c60154546102e6525f68cb4f31e0657                          0.000000   \n",
       "\n",
       "dest_district_hash                            2301bc920194c95cf0c7486e5675243c  \\\n",
       "time_bucket start_district_hash                                                  \n",
       "2016-01-20  08232402614a9b48895cc3d0aeb0e9f2                          0.004906   \n",
       "            08f5b445ec6b29deba62e6fd8b0325a6                          0.000000   \n",
       "            0a5fef95db34383403d11cb6af937309                          0.000000   \n",
       "            1afd7afbc81ecc1b13886a569d869e8a                          0.000079   \n",
       "            1c60154546102e6525f68cb4f31e0657                          0.000693   \n",
       "\n",
       "dest_district_hash                            2350be163432e42270d2670cb3c02f80  \\\n",
       "time_bucket start_district_hash                                                  \n",
       "2016-01-20  08232402614a9b48895cc3d0aeb0e9f2                          0.013900   \n",
       "            08f5b445ec6b29deba62e6fd8b0325a6                          0.000000   \n",
       "            0a5fef95db34383403d11cb6af937309                          0.000000   \n",
       "            1afd7afbc81ecc1b13886a569d869e8a                          0.000132   \n",
       "            1c60154546102e6525f68cb4f31e0657                          0.000433   \n",
       "\n",
       "dest_district_hash                            2407d482f0ffa22a947068f2551fe62c  \\\n",
       "time_bucket start_district_hash                                                  \n",
       "2016-01-20  08232402614a9b48895cc3d0aeb0e9f2                          0.001635   \n",
       "            08f5b445ec6b29deba62e6fd8b0325a6                          0.052225   \n",
       "            0a5fef95db34383403d11cb6af937309                          0.000000   \n",
       "            1afd7afbc81ecc1b13886a569d869e8a                          0.207689   \n",
       "            1c60154546102e6525f68cb4f31e0657                          0.000519   \n",
       "\n",
       "dest_district_hash                                          ...                 \\\n",
       "time_bucket start_district_hash                             ...                  \n",
       "2016-01-20  08232402614a9b48895cc3d0aeb0e9f2                ...                  \n",
       "            08f5b445ec6b29deba62e6fd8b0325a6                ...                  \n",
       "            0a5fef95db34383403d11cb6af937309                ...                  \n",
       "            1afd7afbc81ecc1b13886a569d869e8a                ...                  \n",
       "            1c60154546102e6525f68cb4f31e0657                ...                  \n",
       "\n",
       "dest_district_hash                            d4ec2125aff74eded207d2d915ef682f  \\\n",
       "time_bucket start_district_hash                                                  \n",
       "2016-01-20  08232402614a9b48895cc3d0aeb0e9f2                          0.048651   \n",
       "            08f5b445ec6b29deba62e6fd8b0325a6                          0.026885   \n",
       "            0a5fef95db34383403d11cb6af937309                          0.015038   \n",
       "            1afd7afbc81ecc1b13886a569d869e8a                          0.070296   \n",
       "            1c60154546102e6525f68cb4f31e0657                          0.007879   \n",
       "\n",
       "dest_district_hash                            d524868ce69cb9db10fc5af177fb9423  \\\n",
       "time_bucket start_district_hash                                                  \n",
       "2016-01-20  08232402614a9b48895cc3d0aeb0e9f2                          0.025348   \n",
       "            08f5b445ec6b29deba62e6fd8b0325a6                          0.000000   \n",
       "            0a5fef95db34383403d11cb6af937309                          0.000000   \n",
       "            1afd7afbc81ecc1b13886a569d869e8a                          0.000015   \n",
       "            1c60154546102e6525f68cb4f31e0657                          0.000000   \n",
       "\n",
       "dest_district_hash                            d5cb17978de290c56e84c9cf97e63186  \\\n",
       "time_bucket start_district_hash                                                  \n",
       "2016-01-20  08232402614a9b48895cc3d0aeb0e9f2                          0.000000   \n",
       "            08f5b445ec6b29deba62e6fd8b0325a6                          0.000000   \n",
       "            0a5fef95db34383403d11cb6af937309                          0.000000   \n",
       "            1afd7afbc81ecc1b13886a569d869e8a                          0.000151   \n",
       "            1c60154546102e6525f68cb4f31e0657                          0.000000   \n",
       "\n",
       "dest_district_hash                            dd8d3b9665536d6e05b29c2648c0e69a  \\\n",
       "time_bucket start_district_hash                                                  \n",
       "2016-01-20  08232402614a9b48895cc3d0aeb0e9f2                          0.000409   \n",
       "            08f5b445ec6b29deba62e6fd8b0325a6                          0.007417   \n",
       "            0a5fef95db34383403d11cb6af937309                          0.090226   \n",
       "            1afd7afbc81ecc1b13886a569d869e8a                          0.001942   \n",
       "            1c60154546102e6525f68cb4f31e0657                          0.001645   \n",
       "\n",
       "dest_district_hash                            de092beab9305613aca8f79d7d7224e7  \\\n",
       "time_bucket start_district_hash                                                  \n",
       "2016-01-20  08232402614a9b48895cc3d0aeb0e9f2                          0.035568   \n",
       "            08f5b445ec6b29deba62e6fd8b0325a6                          0.000000   \n",
       "            0a5fef95db34383403d11cb6af937309                          0.007519   \n",
       "            1afd7afbc81ecc1b13886a569d869e8a                          0.000004   \n",
       "            1c60154546102e6525f68cb4f31e0657                          0.000000   \n",
       "\n",
       "dest_district_hash                            f2c8c4bb99e6377d21de71275afd6cd2  \\\n",
       "time_bucket start_district_hash                                                  \n",
       "2016-01-20  08232402614a9b48895cc3d0aeb0e9f2                          0.000000   \n",
       "            08f5b445ec6b29deba62e6fd8b0325a6                          0.000000   \n",
       "            0a5fef95db34383403d11cb6af937309                          0.000000   \n",
       "            1afd7afbc81ecc1b13886a569d869e8a                          0.000026   \n",
       "            1c60154546102e6525f68cb4f31e0657                          0.000000   \n",
       "\n",
       "dest_district_hash                            f47f35242ed40655814bc086d7514046  \\\n",
       "time_bucket start_district_hash                                                  \n",
       "2016-01-20  08232402614a9b48895cc3d0aeb0e9f2                          0.002862   \n",
       "            08f5b445ec6b29deba62e6fd8b0325a6                          0.000000   \n",
       "            0a5fef95db34383403d11cb6af937309                          0.000000   \n",
       "            1afd7afbc81ecc1b13886a569d869e8a                          0.000241   \n",
       "            1c60154546102e6525f68cb4f31e0657                          0.000087   \n",
       "\n",
       "dest_district_hash                            f9280c5dab6910ed44e518248048b9fe  \\\n",
       "time_bucket start_district_hash                                                  \n",
       "2016-01-20  08232402614a9b48895cc3d0aeb0e9f2                          0.000000   \n",
       "            08f5b445ec6b29deba62e6fd8b0325a6                          0.000927   \n",
       "            0a5fef95db34383403d11cb6af937309                          0.000000   \n",
       "            1afd7afbc81ecc1b13886a569d869e8a                          0.006696   \n",
       "            1c60154546102e6525f68cb4f31e0657                          0.000000   \n",
       "\n",
       "dest_district_hash                            fc34648599753c9e74ab238e9a4a07ad  \\\n",
       "time_bucket start_district_hash                                                  \n",
       "2016-01-20  08232402614a9b48895cc3d0aeb0e9f2                          0.012265   \n",
       "            08f5b445ec6b29deba62e6fd8b0325a6                          0.000000   \n",
       "            0a5fef95db34383403d11cb6af937309                          0.000000   \n",
       "            1afd7afbc81ecc1b13886a569d869e8a                          0.000094   \n",
       "            1c60154546102e6525f68cb4f31e0657                          0.000952   \n",
       "\n",
       "dest_district_hash                            fff4e8465d1e12621bc361276b6217cf  \n",
       "time_bucket start_district_hash                                                 \n",
       "2016-01-20  08232402614a9b48895cc3d0aeb0e9f2                          0.001226  \n",
       "            08f5b445ec6b29deba62e6fd8b0325a6                          0.000000  \n",
       "            0a5fef95db34383403d11cb6af937309                          0.000000  \n",
       "            1afd7afbc81ecc1b13886a569d869e8a                          0.000004  \n",
       "            1c60154546102e6525f68cb4f31e0657                          0.000000  \n",
       "\n",
       "[5 rows x 66 columns]"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "posterior.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 465,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "Y = np.load('ignored_assets/Y.npy')\n",
    "Y_total = np.sum(Y,axis=1)\n",
    "num = Y.shape[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 304,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "start2 = pd.to_datetime('2016-01-21 00:00:00')\n",
    "end2 = pd.to_datetime('2016-01-22 00:00:00')\n",
    "mask2 = (orderdf_s['time_bucket'] >= start2) & (orderdf_s['time_bucket'] < end2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 305,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "ptable = pd.pivot_table(orderdf_s[mask2], index='time_bucket', \n",
    "               columns=['start_district_hash'], aggfunc=len, \n",
    "               values='order_id', fill_value=0).fillna(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 496,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x11f9a07d0>]"
      ]
     },
     "execution_count": 496,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYcAAAD8CAYAAACcjGjIAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nrkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11f9a0750>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "order = Y[-288:].sum(axis=0).argsort()\n",
    "cols = ptable.as_matrix().sum(axis=0).argsort()\n",
    "plt.plot(Y[-288:].sum(axis=0)[order])\n",
    "plt.plot(ptable.as_matrix().sum(axis=0)[cols])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 497,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 2, 58, 44, 19,  0, 37,  1, 33, 16, 60, 27, 35, 22, 34, 57, 32, 54,\n",
       "       65, 21,  4, 39, 52, 29, 45, 30,  8, 50, 15, 51, 25,  6, 11, 63, 23,\n",
       "        7, 49,  5, 20, 62, 55, 24, 61, 17, 59, 48, 38, 14, 26, 47, 43, 53,\n",
       "       13, 64, 12, 10, 31,  3, 41, 42, 46, 40, 28,  9, 36, 18, 56])"
      ]
     },
     "execution_count": 497,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cols"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 498,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([60, 41, 10, 48, 18, 36, 58, 63, 44, 12, 43, 55, 37, 54,  1, 31, 23,\n",
       "       25, 49, 57, 53, 45, 64, 51, 46, 47, 52, 62, 40, 11, 34, 17, 29, 20,\n",
       "       42, 39, 59, 33, 22, 24, 32, 50,  6,  5, 21, 56, 27, 61, 15,  4, 65,\n",
       "        7, 38, 28,  8, 19, 35, 16, 14,  3,  9,  2,  0, 26, 13, 30])"
      ]
     },
     "execution_count": 498,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "order.argsort()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 499,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([62, 14, 61, 59, 49, 43, 42, 51, 54, 60,  2, 29,  9, 64, 58, 48, 57,\n",
       "       31,  4, 55, 33, 44, 38, 16, 39, 17, 63, 46, 53, 32, 65, 15, 40, 37,\n",
       "       30, 56,  5, 12, 52, 35, 28,  1, 34, 10,  8, 21, 24, 25,  3, 18, 41,\n",
       "       23, 26, 20, 13, 11, 45, 19,  6, 36,  0, 47, 27,  7, 22, 50])"
      ]
     },
     "execution_count": 499,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "order"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 500,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "sorted_cols = ptable.columns[cols]\n",
    "\n",
    "named_cols = sorted_cols[order.argsort()]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 501,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "lastDayPrediction = pd.DataFrame(trainPredict[-288:].clip(min=0).round(), \n",
    "                                 columns = named_cols,\n",
    "                                 index = ptable.index - pd.Timedelta('1 day')\n",
    "                                )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 502,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th>start_district_hash</th>\n",
       "      <th>90c5a34f06ac86aee0fd70e2adce7d8a</th>\n",
       "      <th>f2c8c4bb99e6377d21de71275afd6cd2</th>\n",
       "      <th>58c7a4888306d8ff3a641d1c0feccbe3</th>\n",
       "      <th>b26a240205c852804ff8758628c0a86a</th>\n",
       "      <th>4b9e4cf2fbdc8281b8a1f9f12b80ce4d</th>\n",
       "      <th>1cbfbdd079ef93e74405c53fcfff8567</th>\n",
       "      <th>929ec6c160e6f52c20a4217c7978f681</th>\n",
       "      <th>82cc4851f9e4faa4e54309f8bb73fd7c</th>\n",
       "      <th>b702e920dcd2765e624dc1ce3a770512</th>\n",
       "      <th>4f4041f7db0c7f69892d9b74c1a7efa1</th>\n",
       "      <th>...</th>\n",
       "      <th>ba32abfc048219e933bee869741da911</th>\n",
       "      <th>cb6041cc08444746caf6039d8b9e43cb</th>\n",
       "      <th>d524868ce69cb9db10fc5af177fb9423</th>\n",
       "      <th>49ac89aa860c27e26c0836cb8dab2df2</th>\n",
       "      <th>de092beab9305613aca8f79d7d7224e7</th>\n",
       "      <th>a735449c5c09df639c35a7d61fad3ee5</th>\n",
       "      <th>0a5fef95db34383403d11cb6af937309</th>\n",
       "      <th>bf44d327f0232325c6d5280926d7b37d</th>\n",
       "      <th>825a21aa308dea206adb49c4b77c7805</th>\n",
       "      <th>1ecbb52d73c522f184a6fc53128b1ea1</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>time_bucket</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2016-01-20 00:00:00</th>\n",
       "      <td>28.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>7.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>24.0</td>\n",
       "      <td>35.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-01-20 00:05:00</th>\n",
       "      <td>27.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>7.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>21.0</td>\n",
       "      <td>33.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-01-20 00:10:00</th>\n",
       "      <td>24.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>23.0</td>\n",
       "      <td>32.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-01-20 00:15:00</th>\n",
       "      <td>23.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>21.0</td>\n",
       "      <td>31.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2016-01-20 00:20:00</th>\n",
       "      <td>21.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>19.0</td>\n",
       "      <td>29.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 66 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "start_district_hash  90c5a34f06ac86aee0fd70e2adce7d8a  \\\n",
       "time_bucket                                             \n",
       "2016-01-20 00:00:00                              28.0   \n",
       "2016-01-20 00:05:00                              27.0   \n",
       "2016-01-20 00:10:00                              24.0   \n",
       "2016-01-20 00:15:00                              23.0   \n",
       "2016-01-20 00:20:00                              21.0   \n",
       "\n",
       "start_district_hash  f2c8c4bb99e6377d21de71275afd6cd2  \\\n",
       "time_bucket                                             \n",
       "2016-01-20 00:00:00                               2.0   \n",
       "2016-01-20 00:05:00                               2.0   \n",
       "2016-01-20 00:10:00                               2.0   \n",
       "2016-01-20 00:15:00                               1.0   \n",
       "2016-01-20 00:20:00                               1.0   \n",
       "\n",
       "start_district_hash  58c7a4888306d8ff3a641d1c0feccbe3  \\\n",
       "time_bucket                                             \n",
       "2016-01-20 00:00:00                               0.0   \n",
       "2016-01-20 00:05:00                               0.0   \n",
       "2016-01-20 00:10:00                               0.0   \n",
       "2016-01-20 00:15:00                               0.0   \n",
       "2016-01-20 00:20:00                               0.0   \n",
       "\n",
       "start_district_hash  b26a240205c852804ff8758628c0a86a  \\\n",
       "time_bucket                                             \n",
       "2016-01-20 00:00:00                               7.0   \n",
       "2016-01-20 00:05:00                               7.0   \n",
       "2016-01-20 00:10:00                               6.0   \n",
       "2016-01-20 00:15:00                               5.0   \n",
       "2016-01-20 00:20:00                               5.0   \n",
       "\n",
       "start_district_hash  4b9e4cf2fbdc8281b8a1f9f12b80ce4d  \\\n",
       "time_bucket                                             \n",
       "2016-01-20 00:00:00                               0.0   \n",
       "2016-01-20 00:05:00                               0.0   \n",
       "2016-01-20 00:10:00                               0.0   \n",
       "2016-01-20 00:15:00                               0.0   \n",
       "2016-01-20 00:20:00                               0.0   \n",
       "\n",
       "start_district_hash  1cbfbdd079ef93e74405c53fcfff8567  \\\n",
       "time_bucket                                             \n",
       "2016-01-20 00:00:00                               1.0   \n",
       "2016-01-20 00:05:00                               1.0   \n",
       "2016-01-20 00:10:00                               1.0   \n",
       "2016-01-20 00:15:00                               1.0   \n",
       "2016-01-20 00:20:00                               1.0   \n",
       "\n",
       "start_district_hash  929ec6c160e6f52c20a4217c7978f681  \\\n",
       "time_bucket                                             \n",
       "2016-01-20 00:00:00                              24.0   \n",
       "2016-01-20 00:05:00                              21.0   \n",
       "2016-01-20 00:10:00                              23.0   \n",
       "2016-01-20 00:15:00                              21.0   \n",
       "2016-01-20 00:20:00                              19.0   \n",
       "\n",
       "start_district_hash  82cc4851f9e4faa4e54309f8bb73fd7c  \\\n",
       "time_bucket                                             \n",
       "2016-01-20 00:00:00                              35.0   \n",
       "2016-01-20 00:05:00                              33.0   \n",
       "2016-01-20 00:10:00                              32.0   \n",
       "2016-01-20 00:15:00                              31.0   \n",
       "2016-01-20 00:20:00                              29.0   \n",
       "\n",
       "start_district_hash  b702e920dcd2765e624dc1ce3a770512  \\\n",
       "time_bucket                                             \n",
       "2016-01-20 00:00:00                               5.0   \n",
       "2016-01-20 00:05:00                               4.0   \n",
       "2016-01-20 00:10:00                               4.0   \n",
       "2016-01-20 00:15:00                               4.0   \n",
       "2016-01-20 00:20:00                               3.0   \n",
       "\n",
       "start_district_hash  4f4041f7db0c7f69892d9b74c1a7efa1  \\\n",
       "time_bucket                                             \n",
       "2016-01-20 00:00:00                               0.0   \n",
       "2016-01-20 00:05:00                               0.0   \n",
       "2016-01-20 00:10:00                               0.0   \n",
       "2016-01-20 00:15:00                               0.0   \n",
       "2016-01-20 00:20:00                               0.0   \n",
       "\n",
       "start_district_hash                ...                 \\\n",
       "time_bucket                        ...                  \n",
       "2016-01-20 00:00:00                ...                  \n",
       "2016-01-20 00:05:00                ...                  \n",
       "2016-01-20 00:10:00                ...                  \n",
       "2016-01-20 00:15:00                ...                  \n",
       "2016-01-20 00:20:00                ...                  \n",
       "\n",
       "start_district_hash  ba32abfc048219e933bee869741da911  \\\n",
       "time_bucket                                             \n",
       "2016-01-20 00:00:00                               1.0   \n",
       "2016-01-20 00:05:00                               1.0   \n",
       "2016-01-20 00:10:00                               1.0   \n",
       "2016-01-20 00:15:00                               1.0   \n",
       "2016-01-20 00:20:00                               1.0   \n",
       "\n",
       "start_district_hash  cb6041cc08444746caf6039d8b9e43cb  \\\n",
       "time_bucket                                             \n",
       "2016-01-20 00:00:00                               0.0   \n",
       "2016-01-20 00:05:00                               0.0   \n",
       "2016-01-20 00:10:00                               0.0   \n",
       "2016-01-20 00:15:00                               0.0   \n",
       "2016-01-20 00:20:00                               0.0   \n",
       "\n",
       "start_district_hash  d524868ce69cb9db10fc5af177fb9423  \\\n",
       "time_bucket                                             \n",
       "2016-01-20 00:00:00                               0.0   \n",
       "2016-01-20 00:05:00                               0.0   \n",
       "2016-01-20 00:10:00                               0.0   \n",
       "2016-01-20 00:15:00                               0.0   \n",
       "2016-01-20 00:20:00                               0.0   \n",
       "\n",
       "start_district_hash  49ac89aa860c27e26c0836cb8dab2df2  \\\n",
       "time_bucket                                             \n",
       "2016-01-20 00:00:00                               0.0   \n",
       "2016-01-20 00:05:00                               0.0   \n",
       "2016-01-20 00:10:00                               0.0   \n",
       "2016-01-20 00:15:00                               0.0   \n",
       "2016-01-20 00:20:00                               0.0   \n",
       "\n",
       "start_district_hash  de092beab9305613aca8f79d7d7224e7  \\\n",
       "time_bucket                                             \n",
       "2016-01-20 00:00:00                               0.0   \n",
       "2016-01-20 00:05:00                               0.0   \n",
       "2016-01-20 00:10:00                               0.0   \n",
       "2016-01-20 00:15:00                               0.0   \n",
       "2016-01-20 00:20:00                               0.0   \n",
       "\n",
       "start_district_hash  a735449c5c09df639c35a7d61fad3ee5  \\\n",
       "time_bucket                                             \n",
       "2016-01-20 00:00:00                               0.0   \n",
       "2016-01-20 00:05:00                               0.0   \n",
       "2016-01-20 00:10:00                               0.0   \n",
       "2016-01-20 00:15:00                               0.0   \n",
       "2016-01-20 00:20:00                               0.0   \n",
       "\n",
       "start_district_hash  0a5fef95db34383403d11cb6af937309  \\\n",
       "time_bucket                                             \n",
       "2016-01-20 00:00:00                               0.0   \n",
       "2016-01-20 00:05:00                               0.0   \n",
       "2016-01-20 00:10:00                               0.0   \n",
       "2016-01-20 00:15:00                               0.0   \n",
       "2016-01-20 00:20:00                               0.0   \n",
       "\n",
       "start_district_hash  bf44d327f0232325c6d5280926d7b37d  \\\n",
       "time_bucket                                             \n",
       "2016-01-20 00:00:00                               1.0   \n",
       "2016-01-20 00:05:00                               1.0   \n",
       "2016-01-20 00:10:00                               1.0   \n",
       "2016-01-20 00:15:00                               1.0   \n",
       "2016-01-20 00:20:00                               1.0   \n",
       "\n",
       "start_district_hash  825a21aa308dea206adb49c4b77c7805  \\\n",
       "time_bucket                                             \n",
       "2016-01-20 00:00:00                               0.0   \n",
       "2016-01-20 00:05:00                               0.0   \n",
       "2016-01-20 00:10:00                               0.0   \n",
       "2016-01-20 00:15:00                               0.0   \n",
       "2016-01-20 00:20:00                               0.0   \n",
       "\n",
       "start_district_hash  1ecbb52d73c522f184a6fc53128b1ea1  \n",
       "time_bucket                                            \n",
       "2016-01-20 00:00:00                               1.0  \n",
       "2016-01-20 00:05:00                               1.0  \n",
       "2016-01-20 00:10:00                               1.0  \n",
       "2016-01-20 00:15:00                               1.0  \n",
       "2016-01-20 00:20:00                               1.0  \n",
       "\n",
       "[5 rows x 66 columns]"
      ]
     },
     "execution_count": 502,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lastDayPrediction.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "reshape s.t. MultiIndex match"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 503,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "ldp_series = lastDayPrediction.stack()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 504,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "prediction = posterior.multiply(ldp_series, axis='index').fillna(0).round()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 507,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "280294.0"
      ]
     },
     "execution_count": 507,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ldp_series.sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 556,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "pt_pred = prediction.sum(axis=1).unstack()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 550,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'d4ec2125aff74eded207d2d915ef682f'"
      ]
     },
     "execution_count": 550,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lastDayPrediction.sum(axis=0).argmax()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 551,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'d4ec2125aff74eded207d2d915ef682f'"
      ]
     },
     "execution_count": 551,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ptable.sum(axis=0).argmax()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 560,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x121c8f7d0>"
      ]
     },
     "execution_count": 560,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAggAAAELCAYAAAC8gH5QAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nbmjWBkZTEwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x122c47d50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "realtot = ptable.sum(axis=0)\n",
    "misspred = (realtot - pt_pred.sum(axis=0)).divide(realtot)\n",
    "misspred.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 561,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.collections.PathCollection at 0x123e4a8d0>"
      ]
     },
     "execution_count": 561,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX8AAAD8CAYAAACfF6SlAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1238370d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter(realtot,misspred)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The prediction is not necessarily good for low demand stations. Which is _ok_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# convert to  input files!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 562,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "FlowsOut = prediction.sum(axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 563,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "260780.0"
      ]
     },
     "execution_count": 563,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "FlowsOut.sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 603,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "flowsout = FlowsOut.unstack()[core_stations]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 605,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "demand lost due to round-off:  -0.0696197563986\n"
     ]
    }
   ],
   "source": [
    "print 'demand lost due to round-off: ', (FlowsOut.sum() -ldp_series.sum())/ ldp_series.sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 606,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "nsteps = int(pd.Timedelta('1 day') / stepsize)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 607,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "stepsize = pd.Timedelta('5 min')\n",
    "horizon = pd.Timedelta('2 hour')\n",
    "start = pd.to_datetime('2016-01-21 00:00:00')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### The difficult one is the flowsin"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 617,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "True\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x120bfadd0>"
      ]
     },
     "execution_count": 617,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP4AAAD8CAYAAABXXhlaAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJztfXuQXNV55+/rnp7pmdE89Bg9kASSQQiMsADzxmwwYJuY\nrL3ldXkdp2KyxRbZLBsnm9QmdrKVSjabVFK1ZcebdVxRYsfejRPbSUxgseMHGAwGDEiAQFgSCCz0\nQNLoMe9Hz0z32T+m+57f93XfngZGLUx/vyqVTvc9995zz7ln+nv+PgkhwOFwtBYyZ3oADoej+fCN\n73C0IHzjOxwtCN/4DkcLwje+w9GC8I3vcLQgfOM7HC2IN7TxReQWEdkrIvtE5JOLNSiHw3F6Ia83\ngEdEsgBeAPAeAIcAPAng50MIP1684TkcjtOBtjdw7pUA9oUQXgYAEfkqgA8CSN347dIR8ugGAIS+\nLnWsmJM4qOlS6k2lGP9QrTnvZNI+sqcvdirpP2ahs50+0LVsv0wcg0zP0MBzut90IfbLd8QDxfRx\no1iMw1sSz+HnsZ9LbVEgyxTmUq+HbJYGR2No08s71xX7tQ1Px/t0dah+LAdmJuM8FLtNP0J2cjb1\nGGheAz0TP2vIijpF5koLngMAxY74TNmJuC5qTgCUcvFzoHdttider21cjyFbiGNQ4ynMqH7oysd+\n/E7N6TUrddF7SMuUmYzjLvblwchOxmvwM2Rmi6iFqdkRzMxNSs2DhDey8dcCOEifDwG4qt4JeXTj\nKrkJAFC4/gp1bHxNHMrSvdNIQ9toPPbf7vpK0v7ja382aYdpff7sxW9L2ryB7GYqdcQxZF84EK93\nzlmqX9jzctKWzXTtkYnUcYfhkaQ9ed35STs3qsfAz1cYiH8cO/cdT72e9NMfPfqjVFq1TJ1zamt/\n0l5+T/z7PH35eapfqSNutO4dcR5Gr92ANPQ+dST1WOiIL/zsyp6kzc8616tf+Nzg2ILnAMD4ub1x\nDI/uj/fs61H9ptfHOZpeEf+QH7k5zv/Kh/Qf+N6fxHvxeEr7D6p+YUtcT/V+HTul+k2882zqF3d+\nfvu+pD38vgvVOf1Pn6j5DPmDI6iFx/Z/qeb3Fm9k4zcEEbkDwB0AkEfXAr0dDkcz8EY2/mEA6+nz\nuvJ3CiGEbQC2AUBP/7pQ+aXv+OaTql/m5ncm7dxz8ReVf60tPn/0xqQ9tWVd0mYRzV5P8vGXZWaT\n/iUv0i9die4726unKdf79prHMvQLbf8i8698/r5n4zkb1iMN7Q/vStpzW89XxwoXrYrjMVJDcv7B\nk+pz70/is/OvPN8HADKdsd8EjZt/UQGgODSctEsXxPkaubBP9Wsfi2Jp9/PHkvbM+uVJO/uE0RBp\nXnj9rHTS8/09SXvu/LORBpaYhs6P70rvriiN9O8ZQxpCPkoDM9dvUcdyo1H0Z6lv6IaNql/f7vhO\nsDQweuMFSXtitba3d5O0w88wQesPADM982pAcbCxLf1GrPpPAtgkIhtFpB3ARwHc8wau53A4moTX\n/YsfQpgTkf8M4DsAsgC+GEJ4ftFG5nA4ThvekI4fQvgWgG8t0lgcDkeTcNqNe4xiThLrPev0AJC7\nb0fSnrw1Wvy79g2pfoPXDyTt/3nWF5P2nR2fSNpWJy+SXsgW65FztBV39SPxXmwBn1ytvSM9B6Jb\nZezseL228dinr2OpOqeigwFAB+nrg+9YovplyVPUeV68RsbYLXjs2Zn4vEuOkL5POjSg9eihX7ws\naS8f1faD4bd1J+2pZfH5uvYZSzlZqdk+YsFjL/V10/dxrFZv5jXs6uuMz2DmgT0uwxfEuSy2q27I\nzsRjwxfTHOVobO36+dbfPRjHQ7r2xFn64suOT8ZjpHsX2/V7c/S6uJ6rH4nfzyyJcze6RbsKMzPx\n2ds3rk3anSe1XSc7U3Yj1vEoq+s21s3hcLyV4Bvf4WhBvO6Q3deD3p614YrL7gSgXTSAdncpV9+V\nF6t+bYPRJXLOP0RR7KU7NyVtG5jDATcM60pLCxThwB57fQ484XNsMA+7rtpffLXmeCzYXSYXaLcm\n35eDS9jtZANI2OXJLqi2A4OqHwdAFclFlh2Z0gM8Ft1LHERkXU3sbmQX49zKeI7sfEGdk13aj1qw\nLljrsqyAg4bsWE9+ILpjWY0768F0dx4/e5FUD0C/D/JKXFt+pwHtymTwfE0t1+/asp30DtC104Ku\nnn74f2Fs+NCCkXv+i+9wtCB84zscLYimWvUZNiKPrfdFFu+feE71myCL/50DX0vavzW4MmlXiXkX\n1I7+G6M4bwDo/uYzSXvuyigOTqzVySk9L0cxfnRjFLk7T5Ilelonrai49LPjWCfX6TBmFouzhSjW\nWvWgOBDF9iLFcHN0l41X52NT50XvSMbE9E+t056GCrqMqM/iJqsOnYfGVT+rclTA4n2VKsN5E3Rf\nqyKWyKrP6o/17IA8JGProyQ8tTGOe3Bcz9fKv3g0jm91FMdLK3U/pabQc3Qc13kFQ1fHsS790as1\n+x29Ul+7fTy+o1laF6s2VN75TKF28o6F/+I7HC0I3/gORwvCN77D0YJoqjuvr+uscPXm/1Dz2PEr\novtm1X0xyc+6htjVd+6TUad78TdiHrPN2RbSt4VIFIpHtJ40ffM7kjbnoVuw7iwj5EojN5jKkYd2\n57H7h/P+AaA0RS5BcjdaF5JyrZGrqp777eRVUa9f8VCcY3ar2fN4vtguAAAdFLHGejznnQM6co8j\n7zjaryq/POWZCgMmb59sIqz/F4f19dpIR2eX4NFr4rxad95sb7QVsQ3Dzut0io3FzlfnrkM1x1Av\n6pEzJ9nFaee4AnfnORyOVPjGdzhaEE0V9S94R0f4q3vm3VBMogEAnzrrX5L2546/O2nfOfCA6sfH\nXroiisWf3v9Y0h4uaffbPSOXoRbG5rTY+Dur7kvaP5g6J2k/MrpJ9bt1aXT7PTQWSRT6slEE7Gub\nVOd8cMnupP3Hx25O2q9OajH7rK4ooo7OxvH9yurvq37DpegG3DkZx/ozdJ9nps9R59zQFd1OzxSi\nO/DZSR3B+NSp+PmX1kWX1iuFFarfE0MbkvZ7BmIC0J39mprq0Fx07/G8XtIRRV9eVwBY16GTsyr4\neP8O9fnPT74raZ/dEaP4DhR0gtLeUa0yVvDK185N2hPX62jLv7nqb5I2z7dFfyau9Q/Go8r5iWU7\nVb/37fpY0v7Y2VFlva33xaT9Hw/cos5Z2xkj93hd7tr8D6rfksz8u3Ll+w5i+85pF/UdDkc1fOM7\nHC2I5lr121eGa1d8BIBOGAF0njxHMnFSDqCt43/2f/8iaf/GhmuStnRoUd9GhVVgmV1VdBzRZtez\nzjaa+MIJNzb3nMHXYKtyFS8eWfw5MYcTkmwSUhpLq/UYsOWco/DsfHF0HEeSBWNRT/NUsEeEoxkB\nbTnn5+NEFUBHD9Z7Jp7LofPj2g5fFud7+Y80PwN7l5i9mN8Ne0xFSx7TzMhjxK235KXRpM0ekUMf\nPVeds3IHJQfRHknjSXxs9G6MzB13Ud/hcFTDN77D0YJobpJOKSRBLpZCSdFUU4CLTbjh4By23rN4\nHwoFdQ7LPRzMYx8+TUzLjWrLO4vtudEooirxdDq9KIh9doYK7inQnHRqMZvFZCnEZ+cUDTEVX1id\n4vNRlV8ex6d4BYyoz/1UgY+8qQZDn5ViSSKz5VBQgVH5mERkA6NUtRs6R9fR0ZwK7eM0itk4J1lT\nIIffPWFx3qgyPCY15/n0YKM05Ma16q2C0Wj+q/gKKnM5YZ+8NvwX3+FoQfjGdzhaEL7xHY4WRFN1\n/NDZnhBwWEIFpsBWHHnGFcc6OkfkscvO+jJKO2M0G9sCrJuPXUCZfEyisJx7bXTeTK8h/aj0GdH6\nXaBEE/Xsq7SrkHXEHLuuTL+ZFH7ATAcl9pjijjlyhbL7zPLddZL+yMkkVfx27OIiPde6P9PcnGxT\n4TUCgAwl1fD7YEuJsQuWr8cuQDv24maiqT5ELslXtW0oDUVTqistYcry4nFSE7slp6ifpeRWtgl6\nhrTyb6UTi1RCS0S+KCKDIrKLvlsmIt8TkRfL/y+tdw2Hw/HmQiOi/pcA3GK++ySA+0MImwDcX/7s\ncDh+SrCgXBBCeEhENpivPwjghnL7ywAeBPDbC94tRLeNdfmU6uQkM6yLqmYfw3eX5urLmH4gsYqv\nkTGiPh9jd5JyQ9Zx56lnt88z3Zi4qSid7XOkgCO/2si1lEmhsrb3sa5VdnE1Om6Gisgz0ZY8f+zK\nrGKUI5dg3XeDxmfddsn3tkoPj6+Q7npkcJSiheKFoHeA75tUxKl1r9cxx2l4vca9VSGEI+X2UQC1\nU58cDsebEm/Yqh/mg/1TA/5F5A4R2S4i22dnJ9K6ORyOJuL1WvWPiciaEMIREVkDYDCtYwhhG4Bt\nwDz1VkV0sVZJLgLZRYkc1RTY0cLL+fSqoo0ZA1vvWbwv7t2n+k1+6Kqk3UvJQbJTW8dB42MLPYt5\nYhJk2DrLiT2c8ANoETekUEcD2sLLkWRscc4Yr8XohiiqryKKKLsW7CVgK7yle+o4Hr0gnExl1TZO\n+pntXVezXyeM94YSV3h80wNa3cjStZkurbhfU5q1bYhj798TE4/G1kdPgBXhmWY8l1K5CAAmibo7\nPWtfRxbaJLUKlu7VqoLycK2JgnVVYdjyXIbsgvk5AF7/L/49AG4rt28DcPfrvI7D4TgDaMSd9/cA\nHgOwWUQOicjtAP4EwHtE5EUAN5c/OxyOnxI0YtX/+ZRDNy3yWBwOR5PQ3Gq5S9aGq7b8MgBNjADo\nqCbWZ7NP/Fj1Ywrsv/38Z5L2L1/5b5O2LR1lSRkqsOWrur7xeNIuXX9p0rYltPp2R3321NboCus8\nGXVELglmwVV5rc6alsFlbQFpOqIqoWVLiRGYlIMj64BqCuvk2qY0FuvATJ7SaFVdpjcPJiKPoVyK\npvIxn8d2lKoSWoTBS6I9aXpzHHfPU/q513w+8vtxNpwlDalXCozB78ryx2mdyG34kw/pWLiBnbVL\nYvU+ul9/UXZrPnr4KxgpHHUiDofDUQ3f+A5HC6KpSToyPZNUjimZarksBi3bGYkuuGotoF02TNWs\neNAM1xkn3HD0VK/l8yPxPvPw00l76QbtxmIRdfk0ue1YpDWRiVU8bWUw9xqg+deUSGnOz2/Xrsh4\nIN7XyntcrbX3rvh8VsxWYjtV+rFJJ6yWsLgbrLh7DleNIXKRvu54viXioCQWlRRjxprGS9huohFL\nVBF4JaKKcmhpPL/3gBar1bwo965OalLuYlo/q3KyG5HVT1aNrGjPIj2/DyVT4bii5oRBJ+JwOBwp\n8I3vcLQgmsu5155LotGs1XVydRRM+/fEY9ai3h6LjqgKN5wDbjny2NrLCTc2Im+CRGEW7+dMFFhm\na6yWUkzJ9Q8mZ5tFUn72HhMFxiK9ivazFN8k6XPkno3CY0ysjn/nl1IU2ESdQpRtpDpMr9A57vkT\nUW3KUaTiJFn7AZ0clFo087j2LOSIf4C9DDM9WpTlpJZOeibLi8dg70TbRLSi549rb8TY26Iqkqf1\nyxtvCUeXdtGaWe8Ie3o42i9D17aqH/M7Zki8HzcRrRWU9jb2W+6/+A5HC8I3vsPRgvCN73C0IJrL\nuTddSCKvcr3aTddzIOpuHKnV87JO5WUXya1Lf5C0X9lFEWGkFwGaI0+RVpgMOo7IY5cd6/SA5ofj\nY4qD3pTQyjIxB3OsmajCbIpqWi+Lj12HnC1oOeh7B9bE80kHzo3qyrJpYJ2+CuTK7K5DiFFiFx5n\nSpp54Ew2zgLs2qf1cI5AVNz+5tnZjci6e2FlfNfqRWjy2vK7AQDstGMykHrl31SZOHpXhm7YqM7p\nJ3fxLEW09j51RPWrRGlmpxfm7gf8F9/haEn4xnc4WhDNjdzLd0A2z4vd1p03dnb8G9R9OIo0oxu1\nS2TFd6NY9NBYrD6qq9bqJAqmwGZ3kqX4PvWBqH5wRF7RUj+TeK+ouzkBpUF3HicnAVpUZHeXTcrh\nZBzmYpu9OD1JZPTsqE710lhtopB1rVUwcZbux0lJTBpSz52XBpuclCWVTrnzSEy36GTx3vDTsfrY\n//SJpH3qwviu5E/oc0YupOrCJ2JCV5U776IowncdiutZLxGN3XE5Iidh9QKAUqEypA6NXrYGtVA8\nukj02g6H460H3/gORwuiuZF7xVJiHc0M6Fz4Nkr1Zs63zpN6iGzN7stGCy/nlFflgxNY5LNUyCy6\nqoQbQ/3MFl4l3pNVud4YGFXU2CSi2jz5tH5zR6OVuc0kbzBy41S1Vln1dVRgGk14+7imn2bxnJNT\n7Npm6lQHjtfSz8pW/dxoj+1eE/xMdm0zpDowH0I9tI/FhBken6o0DB0pyutuKzDxWvPc8bUtt2KG\nnonXJY23QYqN8Wv4L77D0YLwje9wtCCaK+oT8ge19bKvg5IWSJTuslVxSLTua4tWeVWM0VSx4QKW\nKvDFBPBwEgUHxdiEGw7OYes9i3mWupvpnTOFdMs0J+mkFWMEgEBJLG10TqA8divu9g5sSdo8j3Yt\nVMFJuo9NIGFw0FTHca1uqAoyDVRCstdrPxifL2cs6rY4ZgXZNbrGCwvA7CkqtccjlhZc0afx/Buu\nBfawcHBPWz69KCufw0FNwxdoj8jyQSomS9+nrVlmbmG1CvBffIejJeEb3+FoQfjGdzhaEE125xUT\nl8vkdZo7jQkWcpTMwK49QLtiPrgkRs39v6F0AooqHb0M625RpZFIb7YRWJxwY49V0GZ4+pjMo3Dr\nFalj5eQNjljrNi4krrKrEl84UcVw33HEIKfEcKILoHVd1nOn6kTk9Y4QCQbZGSzmaC04iYUTWACt\ny8/QmnPpMMCUr5qO9og5k/TDc1mkJZtbG92iI+cY3f1QtB9k6kQFMo05z7mlAp+89ZKknad3jynk\nZ657pzqH7QL8DF2PGN6/5AKNVU5upJLOehF5QER+LCLPi8ivlb9fJiLfE5EXy/8vXehaDofjzYFG\nRP05AL8ZQng7gKsB3CkibwfwSQD3hxA2Abi//NnhcPwU4DVX0hGRuwH87/K/G6hi7oMhhM31zu3p\nXxcuvf4TAID8fc+qY0xlzK45W7WEce7nIgHfnk9dlLSzJlJM5aiTK8YmvtjzKqhX7YbFZ/7eJt+o\nqjPffDJp21x/RirFNLTYx5Fx/AxWLJ5bGcVVPt+uBbvCJi6KbSteMnh8YyaRhvP42Y3F42Eab0An\n/fA82CQkFpPTquoAWo07/LNEU01LO/CUFuHtGlZQMJGJ3I9d0TaRhhN4+HnHbozJZsPnaU5BHhO7\n8FjFAaLatfP+z2J86ODiVtIRkQ0ALgXwOIBVIYQKG8BRAKtSTnM4HG8yNLzxRWQJgH8C8OshBBXJ\nEebFhpqig4jcISLbRWT77MxErS4Oh6PJaMiqLyI5zG/6r4QQvlH++piIrCFRf7DWuSGEbQC2AUBv\nz9pQEYczJmpu8B1RFF5For4tbNnz/T1J+9VJbY1OBUWfsTXc0lmNXrshaXOUmqXAZksri3lVCTcp\nSMvnB4AsWY9ZLbGiq6JuSrGO26KZHBXG1YpgrP9sle88FMV2K2azKqGTkrSoz+pQlsbH33cP6rUU\nEplLJPZbtauLxHseQ5vxgnAk4HKiaRs6P3pvbDQci9M839VFRuM72jmS/uPGFXiKKUVClz+v36G0\nqMDiRdoLUmwvS/cN/pQ3YtUXAF8AsDuE8Gk6dA+A28rt2wDc3dgtHQ7HmUYjv/jXAfhFAM+JyDPl\n734HwJ8A+LqI3A7gFQAfOT1DdDgci40FN34I4Yeorr9YwU2LOxyHw9EMNJdzrxiqIvEqyKYkbVnC\nAc44O6srujqmiGcvYyuvckYYRV3ZLL40YglbqZYpsNOubc9hHVFdy9JAU+Rd24Z4zM4b2xOYgKJN\nVaDVk8rlptjtxJF/gM7cY10bhiRCRayRTt0+pl1N7G7UZBJIhSqBReNLI6AAoDLoipaIozNvewMA\n5igY0VJ8q/eBxm1tOcr+oohUDI8gVfAtpZBq2AxBngc+v+ra5fOciMPhcKTCN77D0YJoqqhfassk\nro/2h3epY53kOmHXS7agk2/YDTg6G2mSZxWFtonaSkka4egwC44YtK40Bieq1OP9S0uwEEPqwOK9\nqtJ75cWq3yy5kHJGRK3A/lXnCDp2+7W/qOeH53/m+kjeYd1dpf2x2jCvSz06bcVJR8k3tjoNRw8K\nq12Wq5Fcmez2qzJK0TvANOEZ0oasipgGy9mXdp6lkO8k9Y9VrVmi2p5apueug4leKNrPqgSVOQ+y\nYNDe/P0b6uVwON5S8I3vcLQgXnOSzhtBX8fqcO3aXwCgEzQALaqrpAcW86BF8N/76peT9h9eeE08\nx1pwV2n66OSeRmRTIi+JZVPn6fM54k9X8CFR3ySdsEhfT41g67069sRzql92M3GxEe8c54CzFRgA\nZjZFUZiTVixvHVM8p3lhbL+0uQPSi1kqy71ZI/ZIqAhEwz1YTMnvt+8Xz+WJi6NqNHxxFNNXPqYT\nZFZ8N86lUj8tnx8/H6tuZh5UwhPzO9Kzvvp+nTi2cntcJ34Gq55V7vXo4a9gpHB0cZN0HA7HWwO+\n8R2OFoRvfIejBXHGOPcKF2k9aeScqGeuoEiv4oDWedilNFyKrh12J9kSR8zZxq4XS1TB+np+e+TF\n79QU+SriL61qbbB2BdLj2LVnI/o4KoxddnnS6QHN268y+mgeSkZ3H9oc77v84afjeAwHoCL2IJvB\nBHHGAYb0g763kYAl0re121VnmDF4bZgTMGvsMpyt10tZnazvA1rn73slzvHI5qjXc2QjoN8HrhVQ\nNFx6xStjleWcyeRkcLQdPxO/Qzw2QJPSsI0mjaAmDGZrfm/hv/gORwvCN77D0YJorqifzSbunKok\ngxniriNxqWion9mds3PynKTNLikpaDcKu6TSklvmUdvtZ/spl01a1VrjymHxl0VpK5KqKrYUkWdd\nbtm0Kr2kYthkEk6E4vPtWrA6xK7RKi5DOo/VH1vxVY0hhdfQug11Nd/aqoIdA7vcqu5LLsEcucXa\nJmM7f6Kx8l7WTVqkNrvmrEtRqQGksnA/W1mYnyk3mF7xN4ncy3jknsPhSIFvfIejBdFcUT+UqqqQ\nVLDkCImNJNYqqzl0JZafWfJA0n5oTxS/itDIdKxHLVjKanUvEufZmgpoum7moWtbFS2/tpoMV7jJ\n9sZzbAWZtr7alXRtVRa23rN4z9Z+W82n+9UohnIEnU0yUZZ3urb1QHC/9nxKFB9MheKUijSlVabC\nLvVTXHoHNLUjrw1H1FnPwiypH+zdmO2Pbwsn7wDAiocOx2dgdc8kVqVFLdqxTrwzrkfnofGkzdGp\ngx86V52zcjRy8802ELnXaKKR/+I7HC0I3/gORwuiuaJ+W1si0tngGZDIy6JP6NPWSxb7npmOVn0V\nwGMop4qUN87IWFppppKqeUb5WP/CtN4lQ/3ExRT52S0FNo9d/VU2lmQOzmHrPYv3Kp8fmtqaRWEr\nkmZpzpmOynogsiO1C4ba4BJes0BeGUXxZVSjtNQxGxhV9R5Vrmc8Gjn+TKJ+bjgGvHSebExMtu8k\nrxnzCgRDoc2qEj/vXEpRV0CL7orLwKhGlQQeD+BxOByp8I3vcLQgfOM7HC2Ipur4c11ZnNo6r6v2\n/kS7RLjq6RTxvFl33ol/tTZp39AVuev+af17k7blI8uRbst8cKMbtI668uF4r6Gro5toYrW+Xu9A\nrII6enbUqXLjeeqzRZ3D/Gtdh2Kby1oBOlGEOfJsxBu7pDgij112WeMqzFBizqmPR+KSZTtVN4yn\ncMCtui+9xBRAnIlm/vPkhuL4Q7ZTzJmx8nyx68sSl2QK8RojF1I0Y7u20vC8nro4tkN31KGHNun3\nofv52K5XDbj/6cj9yByF0yt0tOXMkjimlSPRVsHXY2KQ+XFH2077eO13A4jvfGhbpMg9EcmLyBMi\nslNEnheRPyh/v1FEHheRfSLyNRGpbelxOBxvOjQi6hcA3BhC2ArgEgC3iMjVAP4UwGdCCOcBGAJw\n++kbpsPhWEy8Js49EekC8EMAvwLgmwBWhxDmROQaAL8fQnhfvfP72gbCNb0fBABMX67zy1nsXv5s\nFO1sUgaL/h/7ziNJ+ysfeHccp8nHZ/eSSkAxlU2ZW4/pv+tyrKVwyNVz+TH3Gouxdkwc1WfzvFPz\n8alto9cqahYA9P+fx5L27M3vVP1YrWD1YPJDV6X2695B1M8pribbrgd209noRoZSoerQlvM7wO/U\n/n8dxfG1P9B7gROAmAfSRlE2mo/PrlEeA3NMaPVJz2taDj7j8V1/idHxw4vDuSci2XLBzEEA3wPw\nEoDhEEJlZg4BWJt2vsPheHOhoY0fQiiGEC4BsA7AlQAuaPQGInKHiGwXke0zYWrhExwOx2nHa7Lq\nhxCGReQBANcA6BeRtvKv/joAh1PO2QZgGwD09K4LFRHfVtJZTskIHEmWMWIj5y4/O0nVW7iajKks\nIzujCJihCLiq5Buix+aoq4mBvOkXRU+mfsqNRlXBVp1hqqX8fc/GAyZ6kMVzTsSw4jPTZaUVd7QR\neWy9L5B4n7tvh+rH0X8TdJ+e7+9R/Vic5vFNrdOeCh4fJzhxJSMrPgdOQiKPj42Gy98XjxXpmFUp\nWAQ/em2MvMvTFOVGdQIZryFH582QaG+vzf0mDL0cR+7x+z958zuS9tAm7QnIFIgOjsYzvd7Shy++\nVX9ARPrL7U4A7wGwG8ADAD5c7nYbgLsbuqPD4TjjaOQXfw2AL4tIFvN/KL4eQrhXRH4M4Ksi8j8A\nPA3gC6dxnA6HYxGx4MYPITwL4NIa37+MeX3f4XD8lKG52XmZqIvYMlfDFL3U/0LUhay+yKWHnjoV\n9cBZ0sMtb1mnyWyrwPK88b1YH7OcdIz8ccqmK9QmEwF0NCG7B6sIO5hLj/jWisYewc/I91UkGiaL\njCPyeNyWsIOz+jLk4rTVhUuklwvp+DM9OkOMI/na6BoFog/vPKbXiB1rnHk5tk5Xy+0ZicdCobat\nA9CVfbulreWVAAAgAElEQVSProhjpWi6enYZtv/YyMRiClW2nYfuHdGgwDYM7pc/pV2K7EbkZ5i9\nSLs4Z5bMj6m0WDq+w+F468E3vsPRgmiqqJ+ZnEkikSau024ZTgbpqUNMwO6SX1r3aNL+20MxScdG\n5LHbrp67CyTqc7Xbtnw6XTSDueVs9VdWUXTkno6uE47co4QPFvkAXeFGUWAz/55xa/Icd30jRuRN\n2Eo6JN6zq2/cRu6ti26o7ucjAUXPy/q1Ytcai/cM61rlyD2OqLRq1xiJvL2P7q95bQCYI1cfJ7gc\nvYb4HU9p8TmNCpyrLAE6CpXdsX27TTUfirzjakqc5GPddIqwg9yIrKrNf64/Zgv/xXc4WhC+8R2O\nFkRTRf1idwdGr90AoFos69rXU30CgK4RHebLYt8rhWidVdVbTCWXNI47pjsGdMIMi282r5pFRaZk\nbh+PYtaSl0b1uEmN4GQSpue2Y2crs/UScAFLFu/YG2E58jiffpTEdhuRx9Z7Fu+7vvG46pelQp4c\nUTlpLO8slrKYzPex6hlfj8+xlOgcAccRdZy8A+h5OXpVpMBuV8V3NDE7qy/83kwaNZXVOMbYRemq\nA0dvjtFaHt+qPQEDvTE6vutQVPds8loSuZd1q77D4UiBb3yHowXhG9/haEE0N3KPYCubTpO+zTqw\nJexgt9YTQxuSNut0NnKPSzUJtTuOG3cX6eHsNrL8ZgzmYq8X4WejvSqwvPCsS3KkVtFkhLG+yPdV\nkXuG955JHvh8S1rBEXnssmOdHtBkILjy4qRZiSKLiONIi9zLmwjGDLnFOJqxYDIluykqkyP3Oo7r\ntWDXbdexuM7MzWftMkwAoghczPvFlWtVpWfD+9fzfLQFCEVvcr/OY+oUHWFJzzBZtpUlY/XIPYfD\nsRB84zscLYimivrZyVn0PnUEAFAyBBQsCrNox8kRgC6v9Z6BSMLwreeji0ZVNgWAlAqt1t3FYPKO\n3AZTbZfKQCmXFI3NqjK9I1G045JJWeOuZL5ATk7JmuqoLMRzxGB73SjDKOrX48jjhBt2abGLDYAS\n7/HEc0lzWeFCfb3p2qqSIrowHIWsbnBCi+Uo5Gg4XrOsSczisledp6KofuyK+N7ZhLC0irbZaZ3c\npaI06d3o261dsNYlG/vRe3ihngfliqZnsGpJJToyO+2Rew6HIwW+8R2OFsQZs+qPXPj66Kc52u7O\n/u8k7W8OR9XBWqk52o9hK+5wBFYgVWTSiIDdlCjExzJkpe44rsVnzrsfu5oTUnSSTvtY7Yo0vU9p\nlYeTQThqkXn6LB0zX68eRx7nh3PCjY3IY+s9i/elnbtVP6b/5khFtoBX8dORWjFNngpbWWn8sljV\nqAsxom7KRLZx9CVXRsqdH8Xsk0P6nVx3KLZVhVzDc8Ai/Bw9H1OTA8AwVS/iikccfTjxbk11nqd3\nl5+h/9t6jqVSnblg1JAU+C++w9GC8I3vcLQgfOM7HC2IJnPuSRKZ1j6mM6E4GkqVcCI3CqD16ENz\nVEV1Kuo22bzlwdf6cQU2CoxdMSA3XVrUnT3Gz5DmwgJ0JKDNIquKOizDuihL1rVW6UeuJusqzKeU\nsrIRh8oWQP0s+QM7Ffl5s8Y1V6SxqxJTNN+5lVpv5ki5HHmuikd0aBvXMtA8e1rH5/dt7Oz4fLMz\n8fm6tTlJZwySGzgYHZ/f0TZE20lmtbZbcMVefifZ9ViqU3qWzy+a96Ey5zLb2G+5/+I7HC0I3/gO\nRwuiqaJ+aMskCQ0cEQZo95QlZWCwGvCDqXPiORTdZev/VnHrlTHbu059VtTWJOJaHjNbhbYWmBsQ\n0NF67JLKmkqwLK6yqG4j21QyDqsYXL33mHZ9cdyYSiyhslZAnUQawzXXZui2k2sbchGVuEKJPW0k\nCrebyEQWp4WThkzl4s5d0efGLkq7ZixaZ2biTJQOx0QtjugDakSAVp7BuEmzNM8s9lseQVbxOMrQ\nluRKAxNxZLbq6MjK+xp21dEVCA3/4pcr5j4tIveWP28UkcdFZJ+IfE1EGrujw+E443gtov6vYb5m\nXgV/CuAzIYTzAAwBuH0xB+ZwOE4fGhL1RWQdgFsB/BGA3xARAXAjgI+Vu3wZwO8D+Hzd6xRDUr1m\npo6IyxZrFokAoNgRxfNLOqKY93eU3IJ8B5+ixEaOsrKRe2wx5nxwa9WvZ7FPAycEcbKLteqzBbtU\nRxzPFvT8JWDuAZv4kpIkYivkpFFg1+vHCTfKcg8o6z2L93NHab6N6MpgNa5IHAWArp7L71BVtRvm\nOSDZtNgfz5lZYqofkaeIKcxtlR62sGc3ExW4mYdp4uDrJJVFqKJTMac5Ihis3nXu0uprpvzOy6z2\nlqWh0V/8PwPwWwAqStByAMPlEtkAcAjA2gav5XA4zjAaKZP9cwAGQwg7Fuqbcv4dIrJdRLbPzE0u\nfILD4TjtaETUvw7AB0Tk/QDyAHoBfBZAv4i0lX/11wE4XOvkEMI2ANsAoLdnbagklGSf+LHqx1Vj\nOGlBTN4+i5SfO/7upM0JKVUFEylpRDqiGtAJfe0SiY1C17CBK1yhhoNf0gKFAK3asNjePZierKSC\nVSwnAIGLf7JlWwyd1dz62upB2KPVCC5gqaoQGW8L02WxWmETbjg4h633LN7bxB4u5MnzYCnI+Hqs\n0tnKQ2xt79hA6/5CFJ+X7jXcCPTucYUj9vjY5wDNiaXh5hx69hhM0Xx1GKZufqfYOzW1RXukKmpr\n6VRjjroFf/FDCJ8KIawLIWwA8FEA3w8h/AKABwB8uNztNgB3N3RHh8NxxvFGAnh+G/OGvn2Y1/m/\nsDhDcjgcpxuvKYAnhPAggAfL7ZcBXLn4Q3I4HKcbTY3ck7lSdHEYnZXdWp2UtDBn9Kks6VDrSCHa\nz1FuI9qNwskSnMQiJgGoxPxtdJ+cqXzL189yhBl9bzn3cuROYreYGL2ZdT8VpWYiARUX23Bt7job\nwchzrMpDmbXg8/g+lnOPXZ7Mkddh+5G9xCa4VMA6PQDM7T9Q81gw9ps0zj12vwFQvIscoTe+PpKO\nWBdg2uawkaBsV2EXIJftAowbmN4p7tfZo0toqXuRm9rakyrJVFK0q14bHqvvcLQgfOM7HC2IM5ak\nYyPRuvpqRyzZnHJdLfVovDZFpYW84bsjMZRFQJtEwW4xvo/N2+cKPHwsN0puq4M6enAmJQqvZBNd\nuIIMqSJWRJ5ezxTksW3ni8H8hexetK5VjpRjvkKbpKPUCjpn2vDdcTQiJ9yoiDyTZ8/iPYv9mT5D\n3U0qGfMkWpcbz8v4ai1OV2AjOdPy8S2XoaJpJzXV8v6pcZNaMkuubOYDBIBuekc5ErBknq+ipgTx\nSjoOhyMFvvEdjhZEc636lKRji2Fy/jSL4FYlKJwXo6E+3h+jiH/4yup4H5OcMkcReZzCMD2gRbFu\nFnnpnBljae3aF8XGmbfVzs1nKz6greMzlK9ux6BosOrkwmdJdeDkDbYCB+ONsOJh0m+rjjAbIxpt\nHo9Ws7Saw2qEpcBmMZ49FcU6RUHZes/ivY3wK11/aexH51TRqtHnQmQwx8zGqN4NDelzuh4hb0md\nCkXsVWGPjZ2HIaJV7yvEd4CjUWev0moEe2KmB6J61vP9PapfpYJSZjo9epThv/gORwvCN77D0YLw\nje9wtCCaquMXO7IYP7cXQLWOoqrOkhvFurtY5/zzk+9K2mwzsHxris+Nop+yRg9kXVdx2s3oaCh2\npaXBVkadPC8qlvn7nk3aXUa/ZijyDqNfs22A54TtI2xXAIBMIY6JdeD8fdqd1zMS3WxjRB7BWZMA\n0M0RluTi4rJW8+OLtgbmyOP5tpx7cylRlKzTA7pMldb39TvA2XrZi2gt9sZ5WPqiIVghGwmTmFj7\nQWeBIjEpOm/UzEP3YSJJoWcaviWdhITB2X0286/iziver93IafBffIejBeEb3+FoQTRV1M9OFND7\n6H4A1aLr8AWxYuuKh6Koz5VgAe3eO7sjirLPkUvEJukoqmZKdunecUD1U2IyibWdhtKZ3TedHL2W\nQscMAF3TUWwskohbFWlH/HSs5qjoMAC95LbjhCBLP83gCsVcbbVo1A12pVXWC6imgeZ+HInGVWsB\nw6dICS31knf4ehyRV0WywuI9if0549JlsX3581Ecf/Vd6VuACVd4/rsMwYlOuorv3pKXanMcAto1\n2rebErNmelU/fkc10UjtqriZ2dqVmKr6NdTL4XC8peAb3+FoQTS3aGY2m5qPXaQAtkBRb1VFJUmc\nPkAU0yyW2RQMa2FPztmvRf3Zd0bxq50s1laEV5GBlLyhCncakXuOxscRdG1GLSnSNTjdwubCW9E/\nuTYn+Rga8GJ7vCJHotmIPitOV2DXouM4Vf2pm5xSu+oP579XceRRMhUn3FiLOlvvWby3RSXb+jmp\nKY67bTxX83tAe3aAeL7dNNMUYdk5kq6mKm4DSpKqW8CU1TjySM32Gur0SpJOmyfpOByOFPjGdzha\nEL7xHY4WRFN1/FIum0S92cyl7Ex057FLCxTxBugSTHuJ4IH1Mauzqgg20sktz1uB2ux2smCdM6Oy\nyKhUlzmHddOufZEr0HLzKa44jlgzFWhZ588qmwgRgxgdnyMQOTLOlnoqUdYcZzZaF5LKBKRnn16h\nbSrtYzEnkrniuKxVxvAfMkeecnlWkaKQbYBcdm02Q5PsObMpPPZ2HoopEZrW3pKGqszL56LdoY1s\nMZVo1lrIE1nJHFUanjVRgRX7Tci6ju9wOFLgG9/haEE0l3MvJ4kYOHS+LgE0fDG7MWKE2Nh6Lbqs\nHSWSDkRReOj86OpoH9eCdnFzrOeZJZ6C/j3jqt/gJVFEXYmoejDJBACMEflG/9MnkvYskWOMbtQi\nKbsrhzZFMXv5rnSxceKseNLMEj0Pfa9EMT5HqsfQZrrvZj2GUxfHecmfiNc+eq12sXYfXUH94n2O\nXqUTQLqO0RwRZbXljRs7O37OzFC1YpoTVdbKXI858gpa81MJNxyRV+WaY/H+m0/G6/3utUl70qiV\n/D50DMVnX3JUV6QdPi+Or+us+K7ZNctRItnJLUSVTe/r8r9+TJ1TvPLiOL4PXVXznqp/Os2fQqNl\nsvcDGMM8gc1cCOFyEVkG4GsANgDYD+AjIYShtGs4HI43D16LqP/uEMIlIYTLy58/CeD+EMImAPeX\nPzscjp8CSAgLV94o/+JfHkI4Qd/tBXBDCOGIiKwB8GAIYXO963RsWBdW/96vAgB6d2mZZHRLlMGX\nUAXT8fM1hxgfayMJfPgy6jer/551HmpMo5neHK3WmUG6z4S+XmFlFCM7Bmtfu9Su53VubfQZdO+K\nInib1iIwR86NDD3S6GYtuspMHFPbZBQpZ/vJgj6sxcHZgSgKy0Qcd35QP1/XsTh2Vg/ah828EiP2\nxPrYL3f+qOo3OxPvVTocIxiL/fGZeF0BYLY2laHiyAN0Pv3ckjiGtnEtZrP1ntWF9X/0aPz+1ivU\nOa/8HEU6zqT/Rob2qJbw+5DdoqMHO/8lWu9Hz6UDG6JnIvNCyoMDWHIwPt/oTTrSMdc+P5f7/+s2\nTO17dUHTfqO/+AHAd0Vkh4jcUf5uVQjhSLl9FEB6WpjD4XhToVHj3rtCCIdFZCWA74mIos8JIQQR\nqSk6lP9Q3AEA2eX9tbo4HI4mo6GNH0I4XP5/UETuwnyV3GMisoZE/cGUc7cB2AYA3SvWh5UPzVtK\n+/eYoJH2aFk+68F4bHBcW5xXbo/HfvJfosCy/P6YKJE1DMPdr0Yxm5NEbDLK0Yl4r94DUWTOH9c5\n8xNro0U2f6KAWrBVWUbOIYqnF+I5TK0M6GQjHl/+lP6jycE4+RPxgdkT0HlSP9/QpniMny83qp+B\nx9R5ajkd0dZspoKaWhd1lJND2lPRTeoMW+tnlkSr+dK9pmISzR/PpaXArqLLKsNa9Tk4h633LN6z\ntR8Alq+5JmlbTxFjZklUqXr3x7m087Dq3peSdue1G5L26LEo3vO6AED+eFxbDkSbXaK9YomHZKK2\ntd9iQVFfRLpFpKfSBvBeALsA3APgtnK32wDc3dAdHQ7HGUcjv/irANwl8zW52gD8XQjh2yLyJICv\ni8jtAF4B8JHTN0yHw7GYWHDjhxBeBrC1xvcnAdx0OgblcDhOLxpy5y0WenvWhisuuxNAtX7NiRis\n5+KJ51S/2ZvfmbT/cNtfJe0/vvHfJO3QkR6+xKQcrJcCmvaaqZ/HTJks5khjHjtORuFEHHtfhh0D\n66Y8R5wQAwBTW7SOV4FNfkqDqoJr7AxMH876dffzuqItV9zlSsMWTJfOpCZMXCIX6Mi9Rs4HoBJz\n+L2ZrVOpliMQp9bFOV6+Q+vHHEWXZZIPwxfJ72h2c4zOK5oK0JNUmoy5DDlh6pVbtU1r9RPRhsGk\nI3adKwQsP9r71xiZXDx3nsPheAvBN77D0YJobrXcuVLiVrGiLye4sOtFVpu4IMrnHi5F0Ynz7GW6\ntosNAKRAyRGGE41540CcdnkjNrLomT/BlWXJj3hMi2IZEhWZF8/muKfx3dm8/Y7jtfkC6lF8s4jK\neexMOQ6Y5+BrGxVKcRFQPn0Vr2LKejD3gNCcVoHX1lSt5Qo3mocwPfmJE26mV8bfPuuyy6Zw+FVR\nolM/pte2jrVcivrB89hhsl3SaLQtEn7FBlV3/8V3OFoQvvEdjhZEc/PxCzMJrdPM9VvUMY446+fK\nKyu12GjpkRIQ9bAVd1nEZbHKXksVaqRKLnkj4haPHKt5TFXwMSIpi6sFsqhbsZopsVn9aTd03cU9\nsaKQUlHovlbkZu9E712x6oytkMNRc/ntke7JFmpkK3N2mkRSc18eR5HmmNeiaOjS2IvB62KhKM2I\nqsy+2EyXxfn0I5vTDeD83iivE1FgAQAoZ57VDatCcdRoWnHT3gOahouvx14Uq3YlqrM49ZbD4UiB\nb3yHowXhG9/haEE0t4RWVx5hy7yeaHXbZaaEUgWsawPANNkG+jN0Drt8DLVylfulDMuxxpF7HEk2\nZuiPWYMduyjqXbnReF8bWcU6GbvSClR+CdDU4KqarLVbkF7O+VztL0Y91+qYih+Qzq9HKz1NPHE2\nGpFtEBxBZ1217OrLkpuTXWSZrReqc5jenPVcOw+dhagrc9XaaTOvDOarC+1xjjnLDoCOGuV3inR6\n2y9Q5J6dB6bb7n3qSNJm9+7JC/UYcqMDqIX2Qm2Xq7vzHA5HKnzjOxwtiOZG7pVCIr5mTKTWBNEf\ndz1C4r1J3mAR+gfjUTxkl5EVcTmKjhNDrDAY6F4snnYZVxNfv+tQFNuVy+eITmhh1YGfvdNGrKVF\nHRr3oIpu5Oclt6YdA7tQOTHHzhfPMYuhFkpFoDmeM1V/2hDFdiX2byYxllxxgF4ncOViEucB7f5k\nd2rVvBKYAnu2O64tk2jMjy+K7XxtW6mYxXvl6jMqAROXMFi96zmoRfW0dbLrUqnMGw419lvuv/gO\nRwvCN77D0YJorlV/bi4R9YZu2KgOVYr+AQAoQswmKXDU2yeWfT1p//DY6tjJiMVsma4Htraz6lAw\nhRpZ1Esr1tmW1+JuoEi70VsvSR0D5+PP9pIYSlZgC45KU4Ust+pIOy5mmY91MZWaBQAzPdGyzNwD\n7MEA9Jr17Y7Xzjz8tOqXoUQrjlhTPHgmKpDXfYrm2HpLRql45JKX4hjmTAIWW9S5wg1TYFuOvNWj\nMZ+ebe1WNVLWexbvDZfEyEevTtoqueu+HXFsl1+rzmFuBF6//m/vVv0qR2Qqxdpv4L/4DkcLwje+\nw9GC8I3vcLQgmsq519O/Llx6/ScAVFegPXpdjKI761uHkvbQ1dp9s/TBnyTt4S9Fl0ZmG2e8aTIL\nRTqR4v4BgBPvjXo5V9K1kX8cscZcc3ztenxr7AK0kYk60y665qwezs+Yxhlv7SMn3xH5/ZY/a2p3\nEdIy46ytgyP5WM89tbVeDQDKPiQbhnV1pc2l5T/sPhxdcLwW1l3MEX9s8xndEHX/Vfe+rM4ZJe57\nnm/OsgO0/YCfg/kYAaDnqz9K2qXrL03aXKeBeRsBzXPIz3DyAyajsjyEPXd9BpPHDzrnnsPhqIZv\nfIejBdFcd14pkjdwBBcArH6k9ilLf6Qjutgd9LGz70va9770rtTbCkWFMWebpahe/jglkJB4aZN5\nWIQepwQeFget2ylPrr7sCwfifYzLrcQ8diSu2gQZdvOAVA9WocREw60cieoQn9/+8C7VL2xYn7RZ\nvOckJgCQNbXrpHa/ms7hx6pNJ51fjzabz+kraDcpPyMTZ8yt167HNk6E2UL02quiGsJlrQBNgc3g\ndxBId7Wyyw7Q4j27PPNEGX/8sg51TqYQ56HjeHwGVkWB6FbOFhYxSUdE+kXkH0Vkj4jsFpFrRGSZ\niHxPRF4s/7904Ss5HI43AxoV9T8L4NshhAswX1VnN4BPArg/hLAJwP3lzw6H46cAC1r1RaQPwDMA\n3haos4jsBXADVct9MISwud61+toGwjW9HwQAjN14gTo2syT+DeIqr9YyzZFy93zpc0n7313+wdT7\nskjPFtmiqWibZv1XYjWAzl3R68DXZpHW0mRz5N5EkyL3LFcdW8Q5N9+qMhy5x/3qR+5RzvwebR1n\nTwWLyWyFtxWFeN15zW3Vn+FLV9QcQ73Ivallcd1Hb4pr3vm49hisfjR6SzIpKhigE2Y4etNG7o2p\nyL34fnHk3uB/0pF7K56LHiWVz2/UkFDmPHxs9G6MzB1fFKv+RgDHAfyNiDwtIn9drpq7KoRQeRuP\nYr64psPh+ClAIxu/DcBlAD4fQrgUwASMWF+WBGqKDiJyh4hsF5HtM6E2E47D4WguGhH1VwP4UQhh\nQ/nz9Zjf+OfhNYr6S5avD1ve9+sAgInV+m/O6JYoJq98KAaDjJyrpZazHo4BG2f995eS9nNfJyop\nUxGFRVIOJlm6V6sRB2+OlvyBnTGQwgaXcGBGmng5ulGLmpwYMkdS7fLnZ5GG8TVRbFRJTAD6Xqld\nTHHofG0VZgxfHMXVc+6N8zC0SVNE5U/FY2yhf/V6fe1OSvfvPBXHYNe2lFK/ski3tRVk0q43m04P\ngOW745rZIC5Wm7q+8XjSfuUPomjNaw4Ax7dGlYfH13tA92O6LM6n5zUH9JqdvIgqAJGBfuVfPKrO\nSQv0GTu79m/2y1/+NKaOLEIATwjhKICDIlLZ1DcB+DGAewDcVv7uNgB3L3Qth8Px5kCjfvxfBfAV\nEWkH8DKAf4/5PxpfF5HbAbwC4COnZ4gOh2Ox0dDGDyE8A+DyGoduWtzhOByOZqCpkXvZybnEPdRt\nSmNlZqJ+vWxnVKjaxzW1NUeZre2M+v6xHdFwyG4ioA5ltXE7DQxENxu7S2yl2v7pGNnGXHMZij5b\nPqjJPzgScLYOsQRHsHVQJJoFu+14fCtHYySgdSlmZ6JbrXvHy9RPRzAqtyZFxg30ahds/njU/7kM\nVP682pTQ9WDp1vn5uskFyLq6RfcOiog0a5anaMSipceu9Dmux7D6iXgvdi9azj2mwGZ3Y5UbmNaa\nI/LYxcw6PaAj/PqJ2y9/Ql+7VHZNH5hyem2Hw5EC3/gORwuiqaJ+KZdNxB8r4rZvjJTHLF5mTUQX\nR4E9dSqKzCoKz0RtsRiq6KtTkkwAHfWWoaouADBL18+QCK/UCHM9VdWVEm4sBTY/HyfzzBrKas1d\nF9WmWU4GMjwC7eTm5Oez0XCVisYAMEcVd5hHADD8fsRRyNxwgHah8jVY5eFrzQ8quq6Yb2B6QEcP\nsquVx5DNa9fjHNFeT37oqng+ud/UewKgmKKyzJgEIAbz8dl5yG8nNY4SbjjRy7pt+1Oou2cvukr1\nq5wXsl4t1+FwpMA3vsPRgmgq9VZf55pwzYZfAlCdGMLgSDSrEjAF1b1/+edJ+8Pn/kzSVvRVqM6f\nrsBaiNkqzEUbx03RTE6YYXpnjhaz4jNb/5lK2iYK8TVKdIzHBmhRnb0WXDSzZFQULtDJVnQWuec/\nx+uxpduONY06y/IApBXHZE4Gy43A4+Pnyz2nPTGTKVTs9Z6Ji2YWLqWqRiZJZ90/H0ItcAFUIL3C\njZ0HpsvifHpWcw58/Fx1zsBTUTVNiz4EAOmYV21+VPgXjJZOOvWWw+Gohm98h6MF4Rvf4WhBNJdz\nj8BkD4B2+TCsPsVYkiG3Wie58Iwrx+qmqd/TeaWO9KmpN6akT167cljp4vta902Wj/H48ulZd2n9\n7DOUUubBfs+kKPnjjfXje8mUjpzM9scosyJHUaqxmow+6sfP127Ko6Wtrb2enecKcu1kUzHLqvR1\nImaxa8vgDE3bq0jXZxtEdnoaaSilvCsVnT4ZU6FsC2jQZue/+A5HC8I3vsPRgmiqO09EjmOewefE\nQn1PM1b4GHwMb9ExnBNCWDBLqqkbHwBEZHsIoVaKr4/Bx+BjaBJc1Hc4WhC+8R2OFsSZ2PjbzsA9\nLXwM8/AxzKPlxtB0Hd/hcJx5uKjvcLQgmrrxReQWEdkrIvtEpCm19kTkiyIyKCK76LumFvwUkfUi\n8oCI/FhEnheRX2v2OEQkLyJPiMjO8hj+oPz9RhF5vLwmXyszKZ82iEi2XJHp3jNx//I994vIcyLy\njIhsL3/X7HfijBaibdrGF5EsgM8B+FkAbwfw8yLy9vpnLQq+BOAW812zC37OAfjNEMLbAVwN4M7y\nszdzHAUAN4YQtgK4BMAtInI1gD8F8JkQwnkAhgDcfhrHAAC/hvmiqxU0+/4VvDuEcAm50Jr9TpzZ\nQrQhhKb8A3ANgO/Q508B+FST7r0BwC76vBfAmnJ7DYC9zZqH8j3vBvCeMzUOAF0AngJwFeaDRtpq\nrdFpuO+68gt9I4B7MZ/C0LT70zj2A1hhvmvaWgDoA/ATlG1sZ2IMzRT11wI4SJ8Plb87EzhjBT9F\nZAOASwE83uxxlMXsZwAMAvgegJcADIcQKpkqp3tN/gzAbwGoMK0sb/L9KwgAvisiO0TkjvJ3zVyL\nM9xJoxYAAAHGSURBVF6ItuWNe2H+z2tTXBsisgTAPwH49RCCKsjXjHGEEIohhEsw/8t7JYALFjhl\n0SAiPwdgMISwY8HOpx/vCiFchnm1804R+Vd8sAlr8YYK0S4GmrnxDwOgShRYV/7uTOBYudAnyv8P\nLtD/DUNEcpjf9F8JIXzjTI0DAEIIwwAewLxo3S8ilRzY07km1wH4gIjsB/BVzIv7n23i/ROEEA6X\n/x8EcBfm/wg2cy0OATgUQqjwZ/0j5v8QNG0Mzdz4TwLYVLbitgP4KOYLb54JNLXgp4gIgC8A2B1C\n+PSZGIeIDIhIf7ndiXkbw27M/wH48OkeQwjhUyGEdWG+6vJHAXw/hPALzbp/BSLSLSI9lTaA9wLY\nhSauRXgzFKI93YYUY7x4P4AXMK9b/m6T7vn3AI4AmMX8X9rbMa9b3g/gRQD3AVh2msfwLsyLbc8C\neKb87/3NHAeAdwB4ujyGXQB+r/z92wA8AWAfgH8A0NGENbkBwL1n4v7l++0s/3u+8h6egXfiEgDb\ny+vxzwCWNnMMHrnncLQgWt6453C0InzjOxwtCN/4DkcLwje+w9GC8I3vcLQgfOM7HC0I3/gORwvC\nN77D0YL4/w1mIcnMBOVWAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11b2dae10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "station_locations = pd.read_csv('inferred_locations.csv')\n",
    "D = np.ceil(squareform(pdist(station_locations[['x','y']].as_matrix())) / 5.)\n",
    "np.fill_diagonal(D, 1.)\n",
    "stations = list(station_locations['start_district_hash'])\n",
    "print stations == list(core_stations)\n",
    "plt.imshow(D)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 692,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "travel_times = pd.DataFrame(D)\n",
    "travel_times.columns = core_stations\n",
    "travel_times.index = core_stations\n",
    "travel_times = travel_times * pd.Timedelta('5 min')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 693,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "a = start + travel_times.loc['08232402614a9b48895cc3d0aeb0e9f2'] - pd.Timedelta('1 day')\n",
    "b = core_stations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 697,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# fuck this, we're doing it slow\n",
    "#begin = start\n",
    "#end = begin + horizon\n",
    "#dts = [begin + j*stepsize for j in range(int(horizon / stepsize))]\n",
    "#mindex = pd.MultiIndex.from_product([dts,core_stations])\n",
    "#base = pd.DataFrame(\n",
    "#        start,\n",
    "#        index=mindex,\n",
    "#        columns = core_stations\n",
    "#    )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 723,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "start = pd.to_datetime('2016-01-20 00:00:00')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 724,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# FlowsOut.unstack().loc[start:end - pd.Timedelta('10 hour')]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 763,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Timestamp('2016-01-20 00:00:00')"
      ]
     },
     "execution_count": 763,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "begin"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 780,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def estimate_arrival(series):\n",
    "    arrival_time = series['time_bucket'] + travel_times['start_district_hash']['dest_district_hash']\n",
    "    return arrival_time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 788,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "test = prediction.T.unstack().reset_index()\n",
    "test = test[test[0] > 0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 802,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "test['arrival_time'] = test['time_bucket'] + travel_times.lookup(test['start_district_hash'], test['dest_district_hash'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 856,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "flowsout = np.zeros((nsteps,), dtype=np.object)\n",
    "flowsin = np.zeros((nsteps,), dtype=np.object)\n",
    "columns = core_stations\n",
    "for i in range(nsteps):\n",
    "    t0 = time.time()\n",
    "    begin = start + i*stepsize\n",
    "    end = begin + horizon\n",
    "    index = [begin + j*stepsize for j in range(int(horizon / stepsize))]\n",
    "    base = pd.DataFrame(\n",
    "        np.zeros((len(index),len(columns)),dtype=np.float64),\n",
    "        index=index,\n",
    "        columns = columns\n",
    "    )\n",
    "    mask = (test['time_bucket'] >  begin) & (test['time_bucket'] <=  end)\n",
    "    \n",
    "    if np.sum(mask) > 0:\n",
    "        flowsout[i] = (base + pd.pivot_table(test[mask], index='time_bucket', \n",
    "               columns=['start_district_hash'], aggfunc=np.sum, \n",
    "               values=0, fill_value=0)).fillna(0).as_matrix()\n",
    "    else:\n",
    "        flowsout[i] = base.fillna(0).as_matrix()\n",
    "    \n",
    "    # ok efficiency\n",
    "    \n",
    "    if np.sum(mask) > 0:\n",
    "        flowsin[i] = (base + pd.pivot_table(test[mask], index='arrival_time', \n",
    "               columns=['dest_district_hash'], aggfunc=np.sum, \n",
    "               values=0, fill_value=0)).fillna(0).as_matrix()\n",
    "    else:\n",
    "        flowsin[i] = base.fillna(0).as_matrix()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 859,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(24, 66)"
      ]
     },
     "execution_count": 859,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "flowsout[-1].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 860,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "predictions  ={\n",
    "        'FlowsIn': flowsin,\n",
    "        'FlowsOut': flowsout\n",
    "    }"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 861,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "sio.savemat('lstm_predictions.mat',predictions)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 863,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 0.,  0.,  0., ...,  0.,  0.,  0.],\n",
       "       [ 0.,  0.,  0., ...,  0.,  0.,  0.],\n",
       "       [ 0.,  0.,  0., ...,  1.,  9.,  0.],\n",
       "       ..., \n",
       "       [ 0.,  0.,  0., ...,  0.,  0.,  0.],\n",
       "       [ 0.,  0.,  0., ...,  0.,  0.,  0.],\n",
       "       [ 0.,  0.,  0., ...,  0.,  0.,  0.]])"
      ]
     },
     "execution_count": 863,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "flowsin[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 864,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "perfect = sio.loadmat('perfect_predictions.mat')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 891,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1511.0"
      ]
     },
     "execution_count": 891,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.abs(perfect['FlowsOut'][0][0].sum() - flowsout[0].sum())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "ierrs = []\n",
    "oerrs =[]\n",
    "\n",
    "for i in range(288):\n",
    "    inerr  = (perfect['FlowsOut'][0][i].sum() - flowsout[i].sum()) / perfect['FlowsOut'][0][i].sum()\n",
    "    ouerr = (perfect['FlowsIn'][0][i].sum() - flowsin[i].sum()) / perfect['FlowsIn'][0][i].sum()\n",
    "    ierrs.append(inerr)\n",
    "    oerrs.append(ouerr)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "plt.plot(ierrs)\n",
    "plt.plot(oerrs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "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": "2.7.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}