Traffic Prediction-checkpoint.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\users\\小柒\\appdata\\local\\programs\\python\\python36\\lib\\site-packages\\sklearn\\cross_validation.py:41: DeprecationWarning: This module was deprecated in version 0.18 in favor of the model_selection module into which all the refactored classes and functions are moved. Also note that the interface of the new CV iterators are different from that of this module. This module will be removed in 0.20.\n",
      "  \"This module will be removed in 0.20.\", DeprecationWarning)\n"
     ]
    }
   ],
   "source": [
    "import scipy\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import xgboost as xgb\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.cluster import KMeans\n",
    "from datetime import datetime\n",
    "% matplotlib inline\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# New York City\n",
    "## Clustering\n",
    "### Extract stations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.collections.PathCollection at 0x168575f9b38>"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAcMAAAGfCAYAAAAjyrlSAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nOo/BkIiIOu//A9uuK7KNqzkzAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x16856133c18>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "path = 'Data/NYC/'\n",
    "nyc_df = pd.read_csv(path+'Bike-NYC.csv')\n",
    "nyc_df = nyc_df[nyc_df['tripduration'] >= 60]\n",
    "startgroups = nyc_df.groupby('start station id').groups\n",
    "index = list(map(lambda x:x[0],startgroups.values()))\n",
    "stationcol = ['id', 'name', 'latitude', 'longitude']\n",
    "startstationcol = list(map(lambda x:'start station '+x, stationcol))\n",
    "station = nyc_df.loc[index, startstationcol]\n",
    "station.columns = stationcol\n",
    "station = station.reset_index().drop('index', axis=1)\n",
    "#station.to_csv(path+'stations.csv',index=False)\n",
    "plt.figure(figsize=(7,7))\n",
    "plt.scatter(station['longitude'], station['latitude'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Uniform geographical grid clustering (GC)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "325 325\n",
      "    id                           name   latitude  longitude  gcluster\n",
      "0   72               W 52 St & 11 Ave  40.767272 -73.993929        22\n",
      "1   79       Franklin St & W Broadway  40.719116 -74.006667         9\n",
      "2   82         St James Pl & Pearl St  40.711174 -74.000165        10\n",
      "3   83  Atlantic Ave & Fort Greene Pl  40.683826 -73.976323         1\n",
      "4  116                W 17 St & 8 Ave  40.741776 -74.001497        15\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAbwAAAGfCAYAAAAkiGdHAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\niIi0BAU8ERFpCQp4IiLSEhTwRESkJfz/wGqtTo6fI2oAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x168576a4278>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "cell = (6, 5)\n",
    "lat, lon = station['latitude'], station['longitude']\n",
    "y1, y2, x1, x2 = lat.min(), lat.max(), lon.min(), lon.max()\n",
    "dy, dx = (y2-y1+0.001)/cell[0], (x2-x1+0.001)/cell[1]\n",
    "iy = list(map(lambda i:(lat>=y1+i*dy)&(lat<y1+(i+1)*dy), range(cell[0])))\n",
    "ix = list(map(lambda i:(lon>=x1+i*dx)&(lon<x1+(i+1)*dx), range(cell[1])))\n",
    "\n",
    "_gclusters, cnt, s = [], 0, 0\n",
    "station['gcluster'] = -1\n",
    "for y in iy:\n",
    "    for x in ix:\n",
    "        c = station[y & x]\n",
    "        s += len(c)\n",
    "        if len(c) > 0:\n",
    "            station.loc[y & x, 'gcluster'] = cnt\n",
    "            _gclusters += [c]\n",
    "            cnt += 1\n",
    "print(len(station), s)\n",
    "print(station.head())\n",
    "plt.figure(figsize=(7,7))\n",
    "for c in _gclusters:\n",
    "    plt.scatter(c['longitude'], c['latitude'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Bipartite clustering (BC)\n",
    "#### Initial k-means clustering"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAbwAAAGfCAYAAAAkiGdHAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4wLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvpW3flQAAIABJREFUeJztvX14VfWZ7/29k2ySiJJUhUkA2wCC\niBqLk9FqSkfJiGeEiFXr2HqmHp4+49WZ9hlKixXs1RxKX6BXnRN12qs9Tj2MznTK+FYpYEcdsB0m\nWlsQTRURKuaMQFJQS6hIQoDf88fea7P3zlp7r5ffb73s/f1clxfslb3WXgsw39z3/b3vW5RSIIQQ\nQsqdqqhvgBBCCAkDCh4hhJCKgIJHCCGkIqDgEUIIqQgoeIQQQioCCh4hhJCKgIJHCCGkIqDgEUII\nqQgoeIQQQiqCmqhvwAtnn322amlpifo2CCGExIRt27a9rZQa7+a9iRK8lpYWbN26NerbIIQQEhNE\n5P+6fS9TmoQQQioCCh4hhJCKgIJHCCGkIqDgEUIIqQgoeIQQQioCCh4hhJCKgIJHCCGkIqDgEUII\nqQgoeIQQQioCCh4hhJCKgIJHCCGkIqDgEUIIqQgoeIQQQioCCh4hhJCKgIJHCCGkIqDgEUIIqQgo\neIQQQioCCh4hdvQ+DHRfCKxoTP/a+3DUd0QICUhN1DdASOzofRhY/7fAyNH068G30q8BoPXm6O6L\nEBIIRniEFLJp5Smxsxg5mj5OCEksFDxCChnc6+04ISQRUPAIKaRhsrfjFqz7ERJrKHiEFNLRBaTq\n84+l6tPHnbDqfoNvAVCn6n4UPUJiAwWPkEJabwY67wMazgEg6V877ytuWGHdj5DYQ5cmIXa03uzN\nkcm6HyGxhxEeITrwW/cjhIQGBY8QHfip++mEhhlCSsKUJiE6sNKfm1am05gNk9NiF0ajOhvlCXEF\nBY8QXXit++mimGGGgkdIFqY0CQlCHFKJNMwQ4goKHiF+iUvvHQ0zhLiCgkeIX+LSexe1YYaQhEDB\nI8QvcUkl+mmUJ6QCoWmFlB+9D4fjlmyYnEln2hwPm6gMM4QkCEZ4pLwIs64Ws1Ti4Pr12D23A6+d\nPwu753ZgcP36SO6DkLhCwSPlha66mhv3ZYxSiYPr16P/q104vn8/oBSO79+P/q92UfQIyYEpTVJe\n6KireWnkjkkq8UD3PVBDQ3nH1NAQDnTfg4bOzojuipB4wQiPlBc6LPpxcV964Hh/v6fjhFQiFDxS\nXuioq+l2X4bQnF7T3OzpeFBYLyRJhIJHygsddTWdjdwhmWgmLPkCpK4u75jU1WHCki9o/RyA9UKS\nXCh4pPxovRlY8gqw4lD6V681Np3uyxDTo1U5gieNjWj++koj9bti9UJC4gxNK4QUEmDzgVIKInLq\n9eBeiN0bNTanWxFXnggVCJJOWC8kSYWCR4gdPtyX3c/swuGhEXQtmAURgVIKg2MmoPHY70a/WWNz\netgOzZrm5nQ60+Y4IXGGKU1CNKCUwuGhEbzz/D9jcNV5UCsaMbjqPKx7/yIck9r8N2tuTg874vJb\nL6TRhUQNIzxCNCAi6PrQqxh58QGMOTYMAGg89jt8MnUIqT/+78Dup42NOgs74rKixgPd9+B4fz9q\nmpsxYckXikaThWlXy+iSez1CTCNKqajvwTVtbW1q69atUd8GIfZ0X+gwW/OctHnGEHY1PKmrM2Za\n8cPuuR32ojxxIqZv3hTBHZFyQUS2KaXa3LyXER4hmnAyqDgaVywCDrv2E3GFDY0uJA5Q8Ehs2bhn\nI+598V4MHBlA09gmLL5kMeZPnR/1bdlSzKAyOGYCGgrcm1m8jDErPC9HJBs6utBgKFIaXL8eB7q7\ncbx/ICum4xYssH8eB2h0IXGAphUSSzbu2YgVz61A/5F+KCj0H+nHiudWYOOejVHfmi0iguc+9LlR\nBpVjUovnPvQ5Z3Hw06cX4kaIwfXr8V93fRXH9/fnNZmvWfkDdD+zy/V1wmyMJ8QJCh6JJfe+eC+G\nTuRb7YdODOHeF++N6I5Kc+2ti5H6+HfzprykPv5dXHvrYueT/IwxC7GZ/UB3N2pGhvOOqaEhTH3i\nIRweGoFbD0BDZyeav74SNRMnAiKomTgxVjVGUhkwpUliycCRAU/HSxLSUlgp6N8rmfTzs0Q2xE3r\nx/vt/7wnHB3M9hu6paGzkwJHIoURHoklTWObPB0vSphLYb0OinY7xiz3uuLwv62BTetONbbUxCZP\nYkdIHKDgkViy+JLFqKvOr/nUVddh8SVF0oNOhJUC9COsboZdF15XnRh9HUOb1u1qb0PVKWz500+4\nTmcSEheY0iSxxHJjanFphpUCLCasxdKnpcaY2V0XAKQaUCeNpmjHLViAx17ci6lPPIQJRweRmtiE\n7X/6Caw8+kG8tWGH57QmIVFCwSOxZf7U+XraEPzUyfxgSlidzlcn0xshDCIiGLyiA7+45GNZcTtX\nKby1YQfG1aUCiV263SG+vYOk/KDgkfKnoyu/1w0wkwI0JaxhCbYDS66ekbcFQkQCR3YcNUaigDU8\nUpKNezZi3qPz0PpgK+Y9Oi+2vXCO6FgK6wade/TCuK4HCsUtaBozqp16vb296O7uxooVK9Dd3Y3e\n3l6jn0fiBSM8UhSrAdzqibMawAHEduqJLT7W/fj6DCB4+4NdC0Xnfb6vG8fUYRSjxnp7e7F+/XqM\njIwAAAYHB7E+s7GhtbXV2OeS+EDBI0Up1gDuR/CSNC7MF0GF1WnUWOd9vgZQxzV1GMWosU2bNmXF\nzmJkZASbNm2i4FUIrlOaIlItIttFZEPm9RQReUFEdovIv4rIGJtzbhWRl3L+OykiH858bYyI3C8i\nu0Rkp4jcqO+xiC50NoCbHBemJe3qtYfOBJpbKKJKHZYiilFjg4ODno4DTIGWG15qeIsBvJbz+tsA\nupVS0wH8HsBnCk9QSv1IKfVhpdSHAfwlgD6l1EuZL38FwAGl1AwAswD8ws8DELPobAA3NS5Mi5CG\n2ZxeDM1OTxOpQx2LXKMYNdbQ0ODpuJUCtQTRSoFS9JKLK8ETkckA5gP4Yea1AJgL4NHMWx4EcH2J\ny3wSwI9zXv8/AFYBgFLqpFLqbfe3TcJCZwO49nFhGbQIaYjzKYvi5Lz06ch0ShH6TR1aKdLj+/fn\nDZMuJXp2ItnQ2Ynpmzfh/Nd2YPrmTa7ELkjE1dHRgVQqlXcslUqho6PD9v3FUqAkmbiN8O4B8GUA\nJzOvzwJwSCl1PPN6L4BJJa7xF8gInog0Zo59XUReFJFHROSP3N82CYv5U+djxRUr0Dy2GQJB89hm\nrLhiha+6m9ZxYTloEVK/kZXuNKhmR6bu1KGfFKlfkSwkSMTV29ubFTDLYdrQ0IDOzk7H+p2fFCiJ\nNyUFT0QWIJ163JZ72OatjnOGROQyAO8rpayqew2AyQB6lFKXAHgewN0O594uIltFZOvBgwdL3S4x\nwPyp8/H0TU+j97ZePH3T075NJlrHheWgRUj9RFYm0qCaWyh0pw79pEh11RH9RlyFQqmUykZ2xcwq\nXlOgJP64ifDaAVwnIn0A1iKdyrwHQKOIWC7PyQBGW65OcQvy05nvAHgfwE8yrx8BcIndiUqp+5VS\nbUqptvHjx7u4XRJXdEaLuWgRUj+Rlak0aOvNaUfmikPpXwO2U/hJHTrhJ0Wqq47oN+LyK5ReU6Ak\n/pRsS1BKLQewHABE5EoAS5VSt4rIIwBuQloEbwOwzu58EakC8AkAH8u5phKR9QCuBLAZQAeAHUEe\nhCQDbePCCq4JBJy76aeHLsQ1PXFhwpIv5LU5AKVTpLpaEBoaGmzFrVTE5Vcorehv06ZNGBwcREND\nQ8moMChW6jWsz6s0gvTh3QlgrYh8A8B2AA8AgIhcB6BNKWX9aPwxAHuVUntszv8nEbkHwEEAiwLc\nC6lwtAip1x66iEd+RYEVHXppZPcjkoX09vbi2LFjo467ibj8CiWQFr2wBIeN8eaRJK34aGtrU1u3\nbo36NghJU9gkDqTToCbGliWcINNeCoXAor6+Hn/+539eUgzszk+lUnmGlThEVt3d3Y7CvGTJklDv\nJUmIyDalVJub93LSCiE5PLF9H77z1OvYf+goJjbW445rzsP1sx0MyLpGiVUAQbad29XgAGDMmDGu\nRKlUajIukRVdoeah4BGS4Ynt+7D88d/g6Eh6weq+Q0ex/PHfAEBx0aPAGSWoECil8lKTuZsfgPiM\nHAuSeiXuoOARkuE7T72eFTuLoyMn8J2nXncWPKINp7RiECHofmYXDg+NZNcZKaWwMrPLb8nVMwDE\nJ7Lq6OiwTb3SFaoPrgcioVFYL45b/Xj/IZut4kWOR4WO0V5xo1hTud/2AKUUDg+NYE1PH1Zu2JEV\nuzU9fTg8NJL99xeXfrvW1lZ0dnZmP7dUYzzxDiM8EgrFftKeMW13LDYoTGysxz4bcZvYWG/z7miI\n6/aDoBRLK1qGDa+mEmtRLQCs6enDmp4+AMCi9pa8BbZxiqzCdIVWIhQ8Ypzcn7QBoGvBrOxP2h1t\ne/Evz/1Q+749P2uI7rjmvLwaHgDUp6px1czxaF+92Z2RxTDFppYkWfBKpRX9CoEleta/PQCjtrVH\n0W9HooGCRzzhR0iK/aTdM6x33551j36W1loiluvSvGrmeDy2bZ83I4tBolicGgamDBtWJgEAplS9\njT+u2Yevfe3Xo0SNkVVlwBoecU2QNTy5omfRtWCWkQ0KQbYnXD97EnqWzcWbq+ejZ9lcPLvzoKOR\nJQp0bz+ICybGeOXW7P7HTMGfjd2L06vSzetc9VOZUPCIa4IISe5P2hYrN+wwskFBp4jGzcgSxeLU\nMDBh2BARjKtLYVF7Cxp+/zpX/RCmNIl7/ApJ7k/almHgVA3vRvy++od5Qhp0g0LT2Cb0Hxmd4vMj\norqMLJ4a2ovgZ7RXUjCRVlxy9QwopfC1rz1i+3U2dVcWjPCIa/xGY7k/aVuGga4Fs7CovQUXNlyl\nfYOCzjVEd1xzHupT1XnH6lPVuOOa81xfw2po33foKBRO1QGf2L7P8/0A/rcfhNXOELe2CRGJTesB\niRbO0iSuKTSDAGkhcStQhRMuCl/rvlcv5ppiEVjQ6Kx99WbbKHFSYz16ls31/nA+KGxnANKp0CC7\n8Ux8jqmZlm7maZJk4mWWJgWPeMKPSzPuFI4UA9JR3KobLtLixJyybKPtdmQB8ObqcP7sds/tsF/R\nM3Eipm/WV8cK8jmmRamYmMZheDTxB4dHE2OY2GcXNaZHisWhoT2sdoYgn2N6pqVTjTAuw6OJeVjD\nIxWPaSemjjpgUMJqZwjyOVHNtPS7EZ0kDwoeqXicIi1dEdj1sydh1Q0XYVJjPQTp2p2udKlbwmpn\nCPI5URlL4jI8mpiHKU1S8TiNFNMZgV0/e1KkGxfCamcI8jluZlqaqLVxLU/lQNMKIdDXJ2f6mjpw\n45YNsqE8CKWMJSZMLXRwJhu6NAmJGNPOT7+42g8XUguDV7q7ux0jMWujgl/o0kwudGkSEjFxXCZb\nbGvFovaW9Nc3bMD+ZcuBE/n3bmojw64XBvD8ujfw3rvDOP3MWly+cBpmXGY/yMBkrY3DoysDCh4h\nBojbDE7g1ADvc178D0zt+gZ23HEIf1rfiHOu/zQWLbgWhzdsSO/WKxA7C90tDLteGMCzP9qJ48dO\nAgDee3cYz/5oJwDYih5rbSQoFDxCDBBF752butvhDRtwxU/uz6Yr/+joITT95H4cvmSy7a69XHS3\nMDy/7o2s2FkcP3YSz697w1bw/C5qZbqSWFDwCNHME9v34cjw8VHHTfbeud2EfqC722GBbDeO9zsP\nATfRwvDeu8OejvtZ1MqmcpILBa/CKMfRYIVE6Y60M6sAwAdOS+F/dl5g7D7cbEJXSmFk/wDsppeO\n7B9AamITju+3SVtWVxsxrJx+Zq2tuJ1+Zq3jOV5rbaant5BkwcbzCiLIAtekoHszgVfszCoAcNqY\nGqOi62akl4hg6ANn275v6ANnY8KSJbZN4xNXrzLizrx84TTUjMn/FlQzpgqXL5ym7TPYVE5yoeBV\nEEEWuCaFYu7IMIjKrOJ2pNe0u+6wFbVpd92Bhs5ONH99JWomTgREUDNxotFWhBmXNeGqW2dmI7rT\nz6zFVbfOdHRp+oFrgUguTGlWEDo3gcc1NRq1O7LxtBR+//6I7XGTTFjyBdveucK6W6lJKA2dnaH2\n2s24rEmrwBXi1+hCyhMKXgWhaxN44V48KzUKIBLRy63ZVYnghM0wBR3uSDcTSpzmOASd71DKgell\npFfYohYEL316dvgxupDyhYJXQSy+ZLHtAlevm8CLpUa9CJ6OKLHQJGIndl7ckU6GFzcTSgBg8Ojo\n6K7YcTe4dWAmScjc4LVPz0kc2VROLCh4FYQlJl5E5sj2Azj8VB9OHBpGdWMtxl3ToiU1qitKdDKJ\nVIvgpFKeXJqF4mkZXtxMKLEiPRP9d24cmOWIlz49r+JIKhMKXoXhZYHrke0HcOjx3VAj6W8iJw4N\n49Dju/FH54/HwMiBUe/3khrVFSU61eZOKuV5m7iT4eXup3fhP++8CgCwpqcvK3yL2luyEZ+Fic0L\nYS1vjRte+vS8NrGTyoQuTeLI4af6smJnoUZO4rYD16GuOt/p5zU1qstAo2OX3RPb96F99WbbyAxI\ni6o1liuXQrEDzOy+C2t5a5QUDrFXSjn249kd99rETioTRnjEkROH7L9ZXDlwCT5w3bmB6m+6DDRB\nIyqnRvFcJjbWZ2t2uazcsANdC2Zh3Uv7R9X9epbN9fQcxXDrwAyboIYSC6f6aOO001D33khe5ObU\np+eniZ1UHhQ84kh1Y62t6FU31npKjdqhy0BjRU5+J6s41QAt6lPVWDpvRl7NLreG98aB9/Drvndx\nNBMJW3W/3HsLSljLW72gq2ZWqj56y63n4Zfr9pQU1csXTsu7H0B/EztJPtyHRxwprOEBgKSq0HjD\ndIydPSHw9ePQyzdl2UY4/R8wyYVL89Fte/GHodFzM6tF8Hc3XxyLha8mePCuHseI6rZvtXu6lvVn\naYkeYF8fLYWuiJMkCy6AJdqwc2nqELu44FS7m9RYPyotadeHN3X5k46CGYeFr15xu+n8e5/d7HiN\nz/3AezpXKYUpy5/Mvn5z1bWexI5ULlwAS7QxdvaEshK4QrzUAAu/AYuIYxsCEP3CV6+47fcDgtXM\nCiOxjyycirUH3s17j1UfpegRndClSSqaoK7KO645D/WpasevexlpZrlFpyzbiPbVm0MbeG1RrN+v\nEL+Dn63anyWW7707jGceeg3Pb34Li9pb8Oaqa7GovQVrevqwcsOOUe5NQoLACI/EiihSqNfPnuQ7\nCrPO+9LDLwcaaebU9J77Gabx0u9n1ca81szs+uXkBHAt6vD/ZSI6q/1jXF2KER7RCgWPxAanRncA\nsU6rWoIUpD2i2JaHsASvprkZx/fvtz1uh5/Bz059cdVHT2bFzRI9ih3RDQUvweiIhuJkSnFqdD/8\nVF+sBQ8I3h4R9ZYHIJx+P7e1P4odMQEFL6HoiIbiFlE5Nbo7HXciqo3nQVKjJmZweiWMfj/2y5Eo\noeAlFB3RUNwiqmKN7m4JuxamS1xNzOD0g+mNC35rf4TogIKXUHREQ7oiKl2Mu6bFttF93DUtrq8R\nZi1Mp7gGTYkmCdNLXwlxgoKXUHREQzquoRMrqgxSUwyzFqZbXIOkRDllhJDSUPASio5oyO81TBpd\ngja6h1kLi4PRBOAuOELcwsbzhDJ29gQ03jA9G41VN9Z6nnHp5xqW0cWKDC2jy5Hto/fjRYFdI7ip\nWpjb1USmG8qL7YIjhJyCEV6C0TH2y+s14mJ0cYoyw6yFuTGahGGi4S44QtxBwSOeCGJ00ZUKLdVO\nEaQW5gU34hqGiYa74AhxBwWPeMKv0UVnz19cokygtNFEd53PzpzC3jZC3MEaHvHEuGtaIKn8fzZu\njC7FRMorJtopTNXZ3Nb53GA3eNkyp1x168xsRHf6mbW46taZZWFY2fXCAB68qwff++xmPHhXD3a9\nMBD1LZEEwwiPeMJv64BOkdLdTmGyzqazobyYOeW2b7W7FriktDDQfUp0Q8EjnnFjdCms10l9NdTR\nE6Pe50ekdLRk5GKyzqbTRKPDnJIkESkm8HG7V5IMKHhlRlBjSKnz7bZ+Fw76tavXoVrSCfSc719+\nRUpHg3oupvvpdJlodJhTkiQidJ8S3VDwyoigxpBS53c/swuHh0ayq1uUUli5YQfG1aWw5OoZ2evY\n1etwQqHqtBrImGotIqVzE3scBje7QYc5xa+IRJEGpfuU6IamlTIiqDGk2PlKKRweGsnbRL1yww6s\n6enD4aGRvM3UTnW5k+8fR/OySzF59Rw0L7s0Nit/wmxWD8KMy5oCm1OcxKKYiDiZZUwbSPxuVSfE\nCUZ4ZURQY0ix83M3Ua/p6cOanj4AwKL2llHLOoOYSqLaeA64r7NFtX4ICD542U+UGFUalJsViG4o\neGWEH6HJFRgIADX6Pdb5luhZYgfAdjN1kBmdUe3nc1tnC3v9kG78iEiUtTRuViA6oeAlFLtIyKvQ\nFAqMndjlnm+lMS3+DDXYueI5nDF8Mi8a82sqiVNDuRNhrh8yhVcRYS2NlAsUvATiFAk13jAdjTdM\ndy00tuYSIBvp5Z6fW7Nb1N6CL006Gwcf3YXUsH005sdUoqNXz3RKNC4bEsKEk1xIuUDBSyDFIiEv\nZhBHIVHA5NVz8g6JCMbVpbI1u4Fv/xqpAq0MGo0FbSgPIyUapaMzqoZxt2nQpDS0k8qFgpdAdE0t\n8SowS66eke27MzHeK2hDeRgpUR2TU/wIQ9QN46XSoFHfHyFuYFtCAnESJK9TS/zMxbQMKrruIZeg\nO/5MiHAh18+ehFU3XIRJjfUQAJMa67Hqhotc1+/8WvzjvvMu7vdHCMAIL5Ho3FTupean4x5KEaSh\nXPeMTSeCTE7xa/GP+9SRuN8fIYAHwRORagBbAexTSi0QkSkA1gI4E8CLAP5SKXWs4JxbAdyRc6gV\nwCVKqZdy3vNTAFOVUhf6f4zKwo8LspjRpXnZpaHcg2lMibBF/8A67HnjbgwN96OuthlTpy1Fc9NC\nT9fwIgy5qU+pApSNv6iUU9LNKDgd0MlJkoCXCG8xgNcAjMu8/jaAbqXUWhH5AYDPAPh+7glKqR8B\n+BEAiMhFANYViN0NAN7zf/uVSxw2lesc76UDkyLcP7AOO3d+BSdPpg0rQ8P7sXPnVwDAk+i5FYbC\nmpid2JVySrodBacDOjlJEnBVwxORyQDmA/hh5rUAmAvg0cxbHgRwfYnLfBLAj3OueTqALwL4hrdb\nJn4Io74VJf0D69DTMwe//P0V2DPnS6j+wrtax5fteePurNhZnDx5FHveuNvTddyOy7JLfQKAZE4t\nNVbMyyg4HegYe0aIadxGePcA+DKAMzKvzwJwSCl1PPN6L4BSRY2/AJD7o/DXAfwdgPdd3gPJwWu/\nWVj1LSdM9sfpir6KMTTc7+m4E24t/k6pT3US+NwP5pb8HC+j4LxQzGHKqSgk7pQUPBFZAOCAUmqb\niFxpHbZ5q+OPjCJyGYD3lVKvZF5/GMC5SqklItJS4vNvB3A7AHzwgx8sdbsVgZ9+M9P1Ld3364Vi\n0ZcuwaurbcbQ8H7b415xIww6amJOo+B2/+p3vvrl2HpAko6blGY7gOtEpA9pk8pcpCO+RhGxBHMy\ngNHfDU5xC3LSmQAuB/DHmWv+J4AZIvJzuxOVUvcrpdqUUm3jx493cbvlj5+tCEEs/0e2H0D/6l9h\n77It6F/9KxzZfsD4/XpBV/RVjKnTlqKqKr+5vKqqHlOnLdX2Gbno2BRQOAoOAFb/722+Nx+w9YAk\nnZIRnlJqOYDlAJCJ8JYqpW4VkUcA3IS0CN4GYJ3d+SJSBeATAD6Wc83vI2NwyUR4G5RSV/p/jMrC\nbz3Oj8lER3Rmun6oM/pywooUg7o03RJ0U0DhKLiuBbPStbwn+3Fc5Qup280HbD0gSSdIH96dANaK\nyDcAbAfwAACIyHUA2pRSXZn3fQzAXqXUnkB3SrKEWY/T4e40fb9Tpy3Nq+EBZqKv5qaFrgVORwtD\nkJpY4Sg4K735vY2/s32/G9Fi6wFJOp4ETyn1cwA/z/x+D4BRDVxKqZ8C+GnBOR8pcs0+AOzB80CY\n9Tgd0Znp+w07+ipFGCYaN+SOggPSIniGD9HKNaoUwtYDkiQ4aSWBhNn0rWuZq9RXoypVg5PvHzdy\nv16iL9OEYaJxS6Eb02u/XKFRJRcOiCZJg4KXUMJq+ta1zFUdPQGkqvCBvzgvVs3qJgjDROMXr7VB\np37A08+sxW3fajd6r4TohoJHilKOy1x11NeKEYaJJgheaoM0qpBygoJHShLVMlcT+KmveRXIsEw0\nYUCjCiknuB6IGMHE+iAdeB0RZglkOmJTWYHsH7DtwgGQFs6ZM7+JutqJAAR1tRMxc+Y3Y1Nj9IKO\nfkBC4gIjvHKm92Fg00pgcC/QMBno6AJabw7lo4M4M02mHL3W1/waUOJkovFK4fiwmR9pQt8r73CT\nOUk8FLwk4UXAeh8G1v8tMJL5Zj34Vvo1EIro+a39mbb0e62vhW1A8bMNXffnF44P2/nLAQ6CJmUB\nBS8peBWwTStPvddi5Gj6eEhRnp/an2lLv9f6WpgGFL+zKnWKpN8FtYQkAdbwkkIxAbNjcK+34zHB\ndETltb4W5gxNP7MqLZH0MxvTDroySTnDCC8peBWwhsnpKNDuOGC2vhfg2mHNxXQbLfqZ4uK3BulH\nbHRHZHRlknKGEV5SsITK7fGOLiCVH5kgVZ8+bqVHB98CoE6lR3sfDn6fAa8d9lYCNzQ3LUR7+xZ0\nzP0t2tu3lBQ7r65OCydRKSY2uiMyujJJOUPBSwrFBMyO1puBzvuAhnMASPrXzvvSx72mR70Q8Nph\nWvqtLembNp+Lnp45WVEq3AbuZTt4kM3ofsTGj0gWg5vLSTnDlGZSsFKCXlKFrTfbf91kfU/DtcOw\n9Du5Qf/tlX68f+wELm5Yi+HhftTWNuPlwVtwLHU1llw9o+R1g9Qg/awE8job0w3cXE7KFQpeknAS\nMK+Uqu/F9doacYrEmvEDnEwY/ztRAAAgAElEQVQNYXh4BAAwPLwfU1J/jzdHAKWmjxrGXIhTDRKo\nwqbN55as6XkVG7ciGXW7AyFxgIJXiXR05bc4AM7pUa8GFC/XjhCniCtV9YdRx2qrR3DxaWsh8rmS\n17Vre0hzIvO5+lcFlRJJv+0OhJQbrOFVIsXqe7n4MaC4vXbEeHV9DrtsiyisQQLVo97jtqanCz/t\nDoSUI4zwKhU36VG/zeu6Uq8GcWpAF6nFiROHRr2/1oNA5tYgN20+1/Y9Ya4KYm8dIWkoeMSZmDev\nB5m5adtfN3UpHt++F1NSf4/a6pHse4dPpLBz8Ba052wPd0scVgWZ7q3r7e3Fpk2bMDg4iIaGBnR0\ndKC1tVXLtQnRCQWvnNDdTB5jA4qOmZt2btBjr+zCmyPAxaedcmnuzLg0vYodEI9VQSacnBa9vb1Y\nv349RkbSPyAMDg5i/fr1AEDRI7GDglcumBgWHWMDis6Zm7mR4qWnpSO95uYt2a/7iews/Exq0Y2f\ndge3bNq0KSt2FiMjI9i0aRMFj8QOCl65YGJYtJ/eP7cEjEZ1zdy0jRRf/wogp8TKr9hZxGFVkKne\nusHBQU/HCYkSCl65YKre5seAUkrMNESjumpjQSJFk3v7/BBFLa2hocFW3BoaGox+LiF+YFtCueBU\nV6v/QLj34aaVQcNoM10zN/1GikFmZprAqqVZ4mPV0np7e41+bkdHB1KpVN6xVCqFjo4Oo59LiB8o\neOVCRxdQPWb08eE/+BoK7XuepBsx0zR+TMfMTaeIsFSkGGRmpgmK1dJM0trais7OzmxE19DQgM7O\nTtbvSCxhSrNcaL0Z+NmdwNF384+fHPFcx+t+ZhcOD42ga8EsiAiUUli5YQfG1aVKz5N0I2aa3J86\namN+XZRhb0IvRZS1tNbWVgocSQSM8MqJo7+3P+4hclJK4fDQCNb09GHlhh1ZsVvT04fDQyOlIz03\na4y8bn4wiN9I0W9kaAqnmhlraYScghFeOaEhchIRdC2YBQBY09OHNT19AIBF7S3ZiK8obloZTLo/\nfeAnUoxDf10uHR0def1wAGtphBRCwSsnNPXNWaJniR2AU2JXyoFZTMw0NMbHxRnptb/O9H1bKUVO\nPCHEGfGy3DJq2tra1NatW6O+jXijQVRy05gWi9pb0PWhVyF2gupmOHRhK4KXczMU9swB6ahK94JY\n3eIU1n0TUomIyDalVJur91LwSC65YmelMa3XL41bgsZjvxt9UsM5wJJXil+4+0KHdKuLczP09Mxx\n6L2biPb2LTZneMeEOIVx34RUKl4EjylNkoeIYFxdKq9mZ9X0GrYdsD/JjSlGQytCGM5InSPLLPzc\nt84mcg53JiQNBY+MYsnVM6By5kdaoie/DWCK0WCoCWPzgAlR9XrfOgcyB7kWhZKUG2xLILYUujFF\nJFg7gdO50+el050rGtO/FmmSdztdZdcLA3jwrh5877Ob8eBdPdj1wkDp+8tgot3A61QYnU3kfq8V\n1eQWQkxCwSsHeh92LRqBCLLN3O7ciz8FvPwvrjequ+mZ2/XCAJ790c7s/rf33h3Gsz/a6Vr0dI0s\n83rfuehsIvd7ragmtxBiEqY0k46JtUDFCLLNvPDc7gs9b3go1TP3/Lo38va+AcDxYyfx/Lo3XG0L\nMLXOx0uvn86BzH6vxS0IpBxhhJd0NAxijgwDGx7sNnsXO25Hc9NCtLdvQcfc36K9fYtvsesfWIee\nnjnYtPlc9PTMcT1YWudAZr/X4uQWUo4wwks6ptYChYGBjeqnn1lrK26nn1nr63q7XhjwtTg1yEZ2\nnU3kfq/FyS2kHKHgJR0DohEaBjaqX75wGp790c68tGbNmCpcvnCa52tZ9UDrWlY9EEBJ0XPT3lDM\nBalzILOfa3FyCylHKHhJx4BohIaBmZqWEG15eBeGjhwHAFSn/G0sD1IPLNXeoLP1IBc3rQS5LSd2\nry24BYGUGxS8pGN6ELOGUWVFCWKCKcLxkVNCNXzkhOvILJcg9cBSvXfFXJBBGsxLiWig1U+EJBya\nVsqB1pvT47lWHEr/qlPsSm0vjyHFIjMvONX93NQDS7U3mHBBlmolCLz6iZCEwwiPOFPMAWp4lY9f\nswigx6kJBKsHlmpv0Nl6YFFKRAOvfiIk4VDwiDMROUCDmEUAfU5N67Ms4ZWq/Eix1L0U670z4YJ0\nI6JFVz8RUuYwpUmccbO93ABBU5KXL5yGmjH5/7T9OjVnXNaUvZ7K3JLX6S12tLa2orOzMytGDQ0N\n6OzsHFW/6+3tRXd3N1asWIHu7u6io73c9NxZacxcrPQmIeUOIzziTEQO0KApycLIzGtKtJCg01uc\nKOWC9OrkLNVKUGz1E2Au0uMQahIXKHjEGdMOUAd0pCRnXNYUSIxy0VUT9IofJ2cxES22+mlcXcqY\n2JlovyDEDxQ8UhxDbQPF0Nk8rgPd01vcYsLJ6bj6yUbsdERmJtovCPELBY/4x1CPnu6UpFcKHaIt\nF56Fnb8cCF2ATTg5AYfVTwXoisw4hJrECQoe8YfhLQ06U5JesHOI7vzlAGZ+pAl9r7xTUoCDtFMU\nEuU8S12RmSnRJsQPFLxKJkiEFlGPnk5BscPJoNL3yju47VvtJe8tSDtFIV7mWeo2huiKzDiEmsQJ\nCl6lEjRCi6BHL6iguJkhGcSgYsLN6WaepQljiK7IjEOoSZyg4MWUjXs24t4X78XAkQE0jW3C4ksW\nY/7U+fo+IGiEFsGWhiCC4naGZBCDSpLcnKXQGZlxCDWJCxS8GLJxz0aseG4Fhk4MAQD6j/RjxXMr\nAECf6AWN0Dz06OlKQ/oVlNwZkgDy+s8WtbfkRXpBHKLl5OaMMp1KiCkoeDHk3hfvzYqdxdCJIdz7\n4r36BC9ohOayR09nXcuvoHiZIRnEIRpVO4UpY0hU6VRCTEHBiyEDR+xHVjkd94WOKSouevR01rWc\nBKXlwrPw4F09RQXKywxJvw7RqNopvKQfdUdj7LMjSYKCF0Oaxjah/8joBaJNYzV+4wxpiorOupad\noBT2yDlFkE4zJHWP04qincJt+tFENMY+O5IkKHgxZPEli/NqeABQV12HxZcs1vtBIUxR0V3XKhSU\nB+/qKRlBRjVDMkzcpB9NRGPssyNJgoIXQ6w6nU6XpnHXpwOm61puIkg3MyRN9/cF4bUtz2LL2ofw\nh3fexhlnnY05t3wa58+5yvN1dERjhSnR6dOn4+WXX2afHUkEFLyYMn/qfG2CFIrr0wHTdS23EWSx\nGZJejTVhiuNrW57F0/d/F8ePpZ/xD28fxNP3fxcAPIte0GjMLiX68ssv4+KLL8bu3bvp0iSxh4KX\ncNxEbrpcn36jRK91LS+C4iWCdJoh6cVYo3uaSim2rH0oK3an7m0YW9Y+5FnwgvbWOaVEd+/ejSVL\nlni6F0KigIKXYNxGbjpcn2FFiV4FRUcE6cVYY2o3nhN/eOdtT8eLEXTqCQ0qJOlQ8BKM28hNh+sz\nlN5A+BMUv85IK5J0ws5YE/Y0lTPOOht/ePug7XE/BJl6QoMKSTpVUd8A8Y/byG3xJYtRV12Xd8yr\n6zOU3kCEJyhWJOl0Xae0qJO71NQ0lTm3fBo1Y/KvXTOmFnNu+XTJc1/b8izu/9wi/N0tnbj/c4vw\n2pZnXX+u3bkdHR1IpVJ576NBhSQJCl6CcYrQCo/PnzofK65YgeaxzRAImsc2Y8UVKzxFZm4/Kyhh\nCYpdJJn7WVfdOtM2arx84TTUjMn/3yao67SYMJ0/5yrMu/3zOOPs8YAIzjh7PObd/vmS9TvL7PKH\ntw8CSmXNLm5Ez+nc1OA76OzszEZ0DQ0N6OzspEGFJAamNGOEV1OIl369oK7PsHoDwxrPVSxiLLYG\nSLfr1I0L8/w5V3k2qAQxuxQ79/bvraHAkcTiWvBEpBrAVgD7lFILRGQKgLUAzgTwIoC/VEodKzjn\nVgB35BxqBXAJgF0AHgEwDcAJAOuVUsuCPEjS8WMKMdGv50RYnxXWeC4vDfF2rtFSu/HcotOFmUsQ\ns4tOowwhccJLhLcYwGsAxmVefxtAt1JqrYj8AMBnAHw/9wSl1I8A/AgAROQiAOuUUi+JyGkA7lZK\nPSsiYwBsEpE/V0r9LODzJBa/phCd/XqlCOuzwhjP5TaSNN2GYEpcgphddBtlCIkLrmp4IjIZwHwA\nP8y8FgBzATyaecuDAK4vcZlPAvgxACil3ldKPZv5/TGkI0Rzi9QSQFimEJJmxmVNuOrWmdmIzqlu\nV8w1qgMnEQkqLkHMLkHOJSTOuI3w7gHwZQBnZF6fBeCQUup45vVeAJNKXOMvACwsPCgijQA6Adxr\nd5KI3A7gdgD44Ac/6PJ2k0coA6MDENVoMpO4iSRNu0bn3PLpvBoe4E9c7MaPzbv9875Gklnv0THO\njJA4UVLwRGQBgANKqW0icqV12Oatqsg1LgPwvlLqlYLjNUhHffcppfbYnauUuh/A/QDQ1tbm+BlJ\nR6cpRLc4BWk67x9Yhz1v3I2h4X7U1TZj6rSlaG4a9XNPaHgdC2Z6qasOcXEyvsy7/fO4/XtrfN8X\nBY6UG24ivHYA14nItQDqkK7h3QOgUURqMlHeZAD7i1zjFmTSmQXcD2C3Uuoeb7ddfhQzhXgRMBMT\nUfzWF/sH1mHnzq/g5Mn0zr2h4f3YufMrABCJ6Pmpx4XhGg0qLqaML4SUGyUFTym1HMByAMhEeEuV\nUreKyCMAbkLaqXkbgHV254tIFYBPAPhYwfFvAGgA8P8GuP+yws4U4lXATExE8Vtf3PPG3Vmxszh5\n8ij2vHF3JILnd4qLdW7UmxSctibQVUmIO4L04d0JYG1GuLYDeAAAROQ6AG1KKWt19scA7M1NWWZM\nMF8BsBPAi5khvt9VSv0wwP2UJV4FzIT5xW99cWh49DnFjpvGbz0uiqWuhRTr16OrkhB3eBI8pdTP\nAfw88/s9AC61ec9PAfy04JyPFLxnL+zrgKQArwJmwvzit75YV9uMoeHRme662mbf9xIE0/U4kxRL\nW+oyvnhB144+QsKEo8U0s3HPRsx7dB5aH2zFvEfnYeOejYGu53WkV7G5mX7vze9osqnTlqKqqj7v\nWFVVPaZOW+rqc3VjYixYWBRLW/odP+aXIGPLCIkSjhbTSBDDiJMxpVR0lbvUNPdzCq8FIJCZxU/T\nuVWnC+rS1OX0jFM9ziul0pYzP3qlq/maOqIymmRIUhGlkuP0b2trU1u3bo36NhyZ9+g823Ri89hm\nPH3T047nFQolkBY1K4qyxLD/yABOqzoLX73iS1gwbQGUUli5YQfG1aWw5OoZRu4tagqdnkA6Spw5\n85uRtjfYYTLNV1jDA9Jpy3m3fx4zP3plyX8Hxc73eo9/d0snYPd9QwRfWrve07UICYqIbFNKtbl5\nL1OaGvFrGClmTAHS0dVTNz6Fm878MX736lJs2zElK3ZrevpweGgEpX5wSeokl2JOzzhhOs2Xm7ZU\nAA5Xn473/uT6rNiV+ndQLCrziqnpMEkkyAomEj5MaWrEr2HEjRiJCLoWzAIArOnpw5qePgDAovYW\ndC2YlZfW1HlvURM3p6cTYaT5rH4964edf+jpwz8sfxJA6X8HOlsXojDJxBE3my5IvGCEpxG/i1ZL\nGVMss8nFD12MnuElqBm3PfseN2IX5N6ixsnRGZXT04kwe+FEBJ9oPIDb/uuf8Pk3v4/b/uuf8InG\nA0X/HeiMysI2ycQVnVEzCQdGeBrxu0KnmDHFzghT1/w4hgAcPzwbKzfscCV6Ya4SyiWo4WTqtKW2\nNTwdTk+dY8/C7IXbseVZPPn9+zDuxAgAYNyJ9/Dk9+8DAMxyEB3dURlHj3GNUhKh4GnGj5uxmBjN\ne3TeqPqeVI1g6oz/QHvtx7OpTbeiF+bAZx2jxbw6Pd2KmO6xZ2Gl+ZRS2LjmAVRlxM5CToxg45oH\ncP5Hr7T9d8CB0Pphw3/yoOBFhF0bgp1bslh9r+vGdE1vXF3KVVozbHSNFmtuWujq/V5ETPfYMydB\nAYD7P7dIm8iICKqOHLL9WtWRQ0X/HTAq0wtrmcmDgpdDWCtwvPTrFTObWEaWOIodEL7hxIuImbi3\nQkExZWo44+zx9pHF2eN9X5N4J2jUzGk14UPTSgZLhPqP9ENBZUXIz6SUUhNNSrUh5FLKbBKm2PUP\nrENPzxxs2nwuenrmoH/Adl54lrANJ15ELIx7M2Vq4ILW+HD+nKtw+/fW4Etr1+P2763xJHacVhM+\nFLwMXkSoGG6E00tP3Pyp87Hw3IWokvRfVZVUYeG5C0NfvmqlC9OzMVU2XVhM9MIeLeZFxMK4N1Om\nBrokkw8dntHAlGYGXY3ZbrYbeOmJ27hnI9b9dh1OqvRam5PqJNb9dh1mT5jtSfSCpmv91Lx0jRZz\nixdHZxj3ZtLUwHpcsqHDMxooeBl0NWa7EU4v2wd07LfTsRTWb83LreEkl1ynZW1tM6blCFHh7NDC\nzwLsRczJvenWDONHGKMyNbA2FH/o8IwGpjQz6GrMdrPdwMv2AR2Rp450bVj1uMLU6fDwfuzceRf6\nB9ZlJ4x0P7PL8fzmpoVob9+Cjrm/RXv7lqzYeU3HOt2Pl3OjSD2yNpQMWIeNBkZ4GXQ1ZruN3tz2\nxOmIPHWIpskG8FzsU6dDeOONu3H/r8/Fmp4+LGpvKRrpubumuxaEoO0LYaceuckgGbAvMhooeDno\naMzWOdFk456NOHr86KjjXiNPHaIZVj3OMXU6tD8rdl5bMYK0ICRllqcFa0PJgXXY8KHgGUCHcNqt\nDAKAhjENWH7Zck/X97uxvBA/9TgnnOpiTlvS3xn6AAD3s0NzCbJ5Pcyt7Tpqb6wNEeIMBS9G5Dop\nRSTrzMzltNRp2kaXAek9eWHO1gSKT0SxS50On0jh8d0LAMD17NBcgqRjvZ7rV7R0Nalz+od/kmj2\nSeI9RwkFLyYURnROe8387q8rjDp1ODf9Uqwu1t6+BQDwxht3Y2ioH+8MNeJtWYS1i/8mu/cN8Bbp\nBUnHejk3iGjpqr2xNuQPHT9whC0+XE/kHQpeTLBzUtqha3+djnYHv5Sqi1mp0+5nduHwyZGsuFn7\nAP3MDg2SjnV7bhDR0ll7c1MbYmSQT9AfOKIQHxqUvEPBiwluIjed++ui3IDuti625OoZeW7MuM8O\nDSJaYdbeGBmMJugPHFGIDw1K3mEfXkxwityqpKpkr57OzwtjA7qXsV6F4qZT7LzOBi1FkCWrYfZl\ncazVaLz83RWWG5RSkYiPzqW+lQIFrxi9DwPdFwIrGtO/9j5s7KOcGt+/9dFvofe2Xjx909NanJ/W\nUOv3R95Hqio16vPC2IDe3LQQM2d+E3W1EwEI6monYubMbxobOWZHkIZyJ4KIVphN6owMRuP27677\nmV1YuWFHVvSsYQgY22h7XZPiw+Z17zCl6UTvw8D6vwVGMuaKwbfSrwGg9WbtH2d6I3mhSWXw2CBq\npAaNtY0YHB7U/nmlxnG5qYvp3EheiO59eEBww0hYfVlsXRiNm787pRQOD43kGacsI9VfXTwPp//6\niVDdsTQoeUec3IBxpK2tTW3dujWcD+u+MC1yhTScAyx5JZx70Mi8R+fZNp83j222XTwbhMK2AyCd\nsvQSxQW5hhuh3LT5XAB2//YFHXN/6+oek0phDQ9If3PmxoXSWBGdJXoAssMQdv7nzyk+ESAi25RS\nbW7eywjPicG93o7HnDBNKjqiJ7/XcLv1PMyG8rjByMA/lnEqV/AsIxUnp8QfCp4TDZMdIrzJ4d+L\nBnRtg3CDjnFcfq/hVijDmg0aV/jNuThObRvZml0OXoYhsB0kWmhacaKjC0jlOwmRqk8fTyC6tkG4\nQcdmBb/XcCuUcTDOkHjitHFix5Zns+nMRe0teHPVtVjU3oI1PX15Rhav1+Umi/Cg4DnRejPQeV+6\nZgdJ/9p5nxHDShh4WUkUFB3bxP1ew4tQ2q0SIsSpbeM/1z6EcXWpvAHmXQtmYVF7i6thCGwHiR6m\nNIvRenNiBQ69DwObVqZrjg2TgY4uzG+9OZRZmTo2K/i9RqWnKklwirVtfCnAMAS2g0QPBa8cCbml\nwg4dmxX8XCOsNUakfCnVtuF3GALbQaKHKc24oLPJfdPKU2JnMXI0fbwCYKqSBMFUQzcbxaOHEV4c\n0BGR5aYwbfvLkNiWCkLCxFTbBttBooeN5za1rtDrdkGb3AsF04mENs0TEhfYVhA/2HjulhjUutKf\nG7DJ3S6FWYjVUhEHgSfEMCaEiVsmkk9l1/DiUutyamZ32+ReVBhzWiqAtKAPvgVAnRJ4H/VC3ZsG\nCNGFqX43thUkn8oWvLiMDwva5O4omOcAKw6l05itN2sTeBObBgjRhSlhYltB8qlswQsaWekiaJO7\nW8HUJPDFxncREjWmhIn755JPZQtenMaHtd6cjsRyI7JcirUttN4MXPwpQKrTr6U6/brwGpoEXses\nTEJMYUqY2FaQfCpb8JIyPswy1zjV3nofBl7+F0CdSL9WJ9KvC2tzmgRex6xMQkxhSpjCXNJLzMC2\nhCRQqm3BS1uDBpemjn13hJiE7QOVA9sSEkbubD671yVrb15qcxrmg3J8F4k7XH9E7KDgRUz3M7tw\neGgkO4DW2rc1ri6FJVfPSL+p1G6+CHb36ZiVSQghYVLZNbxcdM6ydIlSCoeHRvL2aVn7tg4PjZza\nr1Wq9hYn8w0hhMQURnhAZBNXrNUiALCmpw9revoAIG/fVt49ONXein2dk1VIwmE9juiCphUg+CzL\ngCilMGX5k9nXb6661vXKkaLYzdhM1cfTiUqIDYXjvIC045LuSGLhxbTClCYQ6cQVK42Zi5XeDExc\nRqcR4hOO8yI6oeABehqyfdQAc2t2i9pb8Oaqa7GovSWvpheothiX0WmE+ITjvIhOKHhAcNNHqcZw\nB0QE4+pSeTW7rgWzsKi9BePqUpDfPBJs2HNcRqcR4hOO8yI6oeABwSeuBEgdLrl6Rp5BxRK9JVfP\nCJ6SpHuTJByO8yI6oUvTIkhDdsDUYaFBJfva8bpvpVOcpVyXpdydhMQcbgknOqHg6cBL47eXNgGn\n6wLIS3EC+ddgKwIpIzg1heiCKU0duE0deq312V23kMIUp896IiGElDsUPB24rQF6rckVXteJ3NQn\nWxEIIcQWpjR14aYG6KfWl3tdxwb5nNQpWxEAAI8NvItVe/qxb3gEk2pTWD61GTc2nRnKZz+xfR++\n89Tr2H/oKCY21uOOa87D9bMnhfLZhBBnKHimsKujBR3y3NFlPzklN3UawSDpuPHYwLtY+vpbOHoy\n3by/d3gES19P/5kEFb1SYvbE9n1Y/vhvcHQkvZtw36GjWP74bwCAokdIxDClaQKnOtr0ecHaBNyk\nTtmKgFV7+rNiZ3H0pMKqPcE2sltitu/QUSicErMntu/Lvuc7T72eFbvsZ4+cwHeeej3QZ9vdS/vq\nzZiybCPaV2/OuwdCiD2M8IJiF8k51dF2P50WqCAOylKpU7YiYN/wiKfjbikmZlb0tv/QUbtTHY/7\ngVEkIf6g4AXBactCodhZDO7VsoC1JGF8RoyZVJvCXhtxm1SbCnRdN2I2sbEe+2ze11CfQvvqzVrq\nem6ElxAyGqY0g+AUyUm1/fuliu0BIbB8ajPqq/JdrfVVguVTmwNdd2KjfYtI7vE7rjkP9an8v/9U\nleDIseNFU6FeCCOKJKQcoeAFwcn5qE7Y98+pE+yJC4Ebm87E3eedg8m1KQiAybUp3H3eOYENK3Zi\nVp+qxh3XnJd9ff3sSVh1w0WY1FgPATCpsR6n19Vg5ERBTTFAXc+N8BJCRsOUZhAcHZHnpOtmP/ls\nWuRysXriKjjlGAY3Np2pvQ3BSheWajm4fvakvGNTlm20vZ7fiOyOa87Lq+EBo4WXEDIaCl4QirUJ\ntN4MPH67/XkV1hOXBNz2zhWKmRuc6np+IzK3wksIyce14IlINYCtAPYppRaIyBQAawGcCeBFAH+p\nlDpWcM6tAO7IOdQK4BKl1Esi8scA/hFAPYAnASxWSVq/DpR2RLInzgi6m8pNux5NRGR+hJeQSkfc\naoyIfBFAG4BxGcF7GMDjSqm1IvIDAC8rpb5f5PyLAKxTSk3NvP4VgMUAfom04N2nlPpZsXtoa2tT\nW7dudXW/saDQxQmkI0Avq4dIHoVN5UDakBKkRte+erNtBDapsR49y+YCCD49hdNXCDGDiGxTSrW5\nea+rCE9EJgOYD+CbAL4o6f01cwF8KvOWBwGsAOAoeAA+CeDHmes1Iy2cz2dePwTgegBFBS9xsCdO\nO8Wayv0KXinXo44IMC4RGYWXVDJuU5r3APgygDMyr88CcEgpdTzzei+AUv/X/AWAhZnfT8qcY+Hm\n/GRS4T1xujHRVF6qxlYufW9sWCeVTsm2BBFZAOCAUmpb7mGbtzrmRkXkMgDvK6Ve8Xq+iNwuIltF\nZOvBgwdL3S4pc5yax4M0lZdqNyiXvrewxp4RElfc9OG1A7hORPqQNqnMRTriaxQRK0KcDGB/kWvc\ngkw6M8PezDkWjucrpe5XSrUppdrGjx/v4nZJOWOiqdyud27VDRdlo55y6XsrF+EmxC8lU5pKqeUA\nlgOAiFwJYKlS6lYReQTATUiL4G0A1tmdLyJVAD4B4GM51+wXkT+IyEcAvADg0wD+PtijkErAqtPp\nXv1TrMZWLn1vutsjCEkaQfrw7gSwVkS+AWA7gAcAQESuA9CmlLLG838MwF6l1J6C8/8ap9oSfoZy\nM6wQY5hoKi9Gkvveck0qDfUppKolb+pLEoWbEL+4bkuIA4lrSyAkQgpNKkB6rufpdTU49P5IooSb\nECe0tyUQQpKHnUll5KTCaWNqsL1rXiT3NLh+PQ5034Pj/f2oaW7GhCVfQENnZ+TXIpUBBY+QmKC7\nRy5uJpXB9evR/9UuqKEhAMDx/fvR/9V05cOrUOm8FqkcuC2BkBjgZpu6V+LmLj3QfU9WoCzU0BAO\ndN8T6bVI5UDBIyQECmkX2CcAABXHSURBVGvlha9N9Mi5WWcUJsf7+z0dD+tapHKg4BFimO5ndmHl\nhh1ZkVNKYeWGHeh+Zlf2PSbSj6X6C8Omptm+V9LpeFjXIpUDBY8QgyilcHhoBGt6+rKit3LDDqzp\n6cPhoZGsCJpKP14/exJ6ls3Fm6vno2fZ3EgdmROWfAFSV5d3TOrqMGHJFyK9FqkcaFohxCAigq4F\nswAAa3r6sKanDwCwqL0FXQtmIT2HPV7N7abcj9Y1dFzb67Xo6CQABY8kDKVUViTsXutGh3PSEj1L\n7ADkiR0Qn+b2JLkfGzo7Xd1Tkp6JmIWN5yQxdD+zC4eHRrJiYaUHx9WlsOTqGdo/z65xuz5V7bkO\nlpvGtCiM8OLC7rkdOL5/9FjbmokTMX3zpkDXLhQeIJ2GbP76Slvh0RWVmXwmEj1eGs9ZwyOJwG0t\nTCc6nJO597movQVvrroWi9pb8p4jTph0Pzq3EnTnH1MqK47H9+8HlMpGZYPr13v+XDo6iQVTmiQR\niAgu/JMm1Jw2gu/LMXz/Z9tQ88Zh/JXBSCmoc9JKh+47dBRja6vROqkhr6Y3ri4VuwivprnZPhrS\n4H50EpiR/QPZ1LT1A0Lnt76Deoc+O69RnslnIsmCER5JBI8NvIulr+/Fe1UARID6Ghy/sBEX/kmT\nMdEI4pzMbSQHgCPDJ3DXT17BE9v3ZUXPbRr2ie370L56M6Ys24j21ZsDNaOXwqT70UlgDtQ3jIra\n637/tu17/URldHQSCwoeSQSr9vTj6MmC9F91Fe589S1jacEgjdul0qFuRdrEBJZiNHR2ovnrK1Ez\ncSIggpqJEx1rbF5xEp49138aa3r6MGX5k9nUb2pik+01/ERlJp8pLB4beBdtz72K5mdfQttzr+Kx\ngXejvqVEwpQmSQT7hkdsj78n6ajARFoziHNSVyN5MeE05eB06370c11gdCvBogULsHL5k9n3dS2Y\nhcNYYmtw8RuVmXqmMEhnN97K/sC3d3gES19/CwBCXZNVDlDwSCKYVJvCXhvRO12J0VpYscWwxdC1\nbDVuA6CDUig8Vhozl/QPMAsA6OnZSzp22Y2jJxVW7emn4HmEgkcSwfKpzXk/5QJAfZXg2+dPxo1N\nZ2n/vKD9d7oayct5S3mhg7Vrway89o2uBQsSKXBHth/A4af6cOLQMKobazHumhaMnT3B9/WcshtO\nx4kzFDySCKyfZFft6ce+4RFMqk1h+dRmIz/hFvbfWXUzAK5FT1cjeZwmsOhGJB2d5/YkxtnB6oYj\n2w/g0OO7oUZOAgBOHBrGocd3A4Bv0XPKbkyqTfm/0QyPDbwbyv9TcYGN54QU0L56s21UNamxHj3L\n5oZ+P7r35PnB5Giuwmk5h366Hgfv8f5ZcRgf1r/6VzhxaHjU8erGWjQvu9TXNQtreEA6u3H3eecE\nEidT1w0bbjwnJABxq5v5rSPqwvRorlyxG1y/HgNd3j8rLuPD7MSu2HE3mMpuVGJtkG0JhBQQt8Wp\nURPmslW/nxWXhbDVjbWejrvlxqYzsfWKC9B/1Yex9YoLtAhSJdYGKXikrPHTv+TUf3fVzPGBG8DD\nbCLXRZijufx+VlzGh427pgWSyv+2KqkqjLumJdT7cINTDVBHbTCuMKVJyha//Ut2hpOrZo7HY9v2\nBTKyeDHDBKnbbdyzEfe+eC8GjgygaWwTFl+yGPOnznd1rh1hjuby+1k67zGIy9J6n06XpimcnM/L\np5bvyDVGeKRsKVajKEXh4tRndx4MPEja7TDqINNVNu7ZiBXPrUD/kX4oKPQf6ceK51Zg456Nru+z\nEF2juQbXr8fuuR147fxZ2D23w3YQtN/P0nWPlsvSqrlZLssj2w+4vsbY2RPQvOxSTF49B83LLo2l\n2AHpH/ruPu8cTK5NQQBMrk0lzrDiFUZ4xBhRW5511ih0GFncXiPIdJV7X7wXQyfya1lDJ4Zw74v3\n+o7ydCxudWsq8ftZupbLHn6qL9tSYKFGTuLwU32xFa4g3Nh0ZlkLXCEUPGKEOIxD0tm/pKMB3O01\ngojrwJEBT8fdEnQ0VzFTSeF1/X6WjvFhJlyWJD4wpUmMECSdaIcf88nyqc2or8pvXvZbowgySNrr\nNYK4RJvG2g9ddjoeFnExlZTClMuSxAMKHjGCznSiFS3uHR6BwqlosZTo6axRXD97ElbdcBEmNdZD\nkG5C97r53O01gojr4ksWo646v5ZVV12HxZcsdn2fJnAyj8RtJ12SXJa5cJuCOzhphRih7blXbdOJ\nk2tT2HrFBZFdKynEyaWpg8IaHpA2lcRxTY/uWZimKZeJKX7xMmmFgkeMoPN/wuZnX4Ldv1IB0H/V\nh4PdqGHiMBYsLsRh9Fc5Uok/EObC0WIkcnSOQzI5PNckOoZQ+yXKKM/ps5O8ky7OVOLEFL9Q8Igx\ndFmek9ogG8XyVuBUL57VnmD14gEwLnpRfnalktQfCKOAphUSe+LWIOt2PFhUQ6iL9eKZJsrPBtL1\nt/7Vv8LeZVvQv/pXnhrGk4pON3K5wwiPJIK4NMh6SVNGtbzVVC9e3D/bxC46k+gyx4S5KzLpMMIj\nxANux4MBenr3/BBlL16Un11sSkrc0DHCLBcT2xTKEQoeIR7wkqbU0bvnxMY9GzHv0XlofbAV8x6d\nlzcrM8pevCg/28SUFFMp0iSJcznBlCYhHvCapjSxvLWUMcQyh5RyaZpwcrr9bBNUN9Y6bhv3g8kU\nKUeYRQMFjxAP3HHNeXk1PMBcmtJJkNwMiM4VPqdrm3JTlvpsU4y7piVPoAB/U1KUUhARo4OkdYsz\ncQcFjxAP2O3KM9FMXkyQdBhD3Iim9Y0/rtiZPhpvmB7ICNL9zC4cHhpB14JZRqMwXeJMvEHBI8Qj\nJtKUhRQTpKaxTeg/MnroshdjSCnRVEph5YYdGFeXwpKrZ3i4c+/4Sa06pRsbb5iO5mWX+roPpRQO\nD41gTU8fAOB2g1FYkhbFlhMUPEJiSDFBWjVnVV70B3g3hjiJZn3VWVmxW9PTh0XtLUYjPb+pVRPp\nRhFB14JZAIA1PX14CzW4E/VIW44y79EYhY2dPYECFzJ0aRISQ4rZ++dPnY8VV6xA89hmCATNY5ux\n4ooVrutmG/dsxNHjo4031RiDd97qwJTlT2bFrmvBrDyxK+YO9YPfRnVT6cZc0ft3HMe3cTQb0VU3\n1qLxhukUqQTDCI+QGLL4ksVFozi/xpDCiMqiYUwDll26DJ//h1PiZid2uo0ufuuRpkwfVnRr8e84\njnMuGIuuBX8SSj3zsYF3tTWQF0bmca/JhgEjPEJiSNAozgm7iAoATkudhm07puQdW7lhB3K3qZgY\nG+a3Ud3E3jqlFP75/2zH/J63sQXj8HzjeHx9ejPW9PSN+rMwgd+9j3Z0P7Mr754tIe9+Zpfmu04W\njPAIiSkm7P1OkVP/kQGs2XoqjWnV8IBTkZ6JsWGlIlknTJg+3n/pIOa8cQQ1mTjgxKFhzD0yAkxv\nxtt1KePR0ao9/XkD0gHg6EmFVXv6PUV5heab3L9P0zXZuEPBI6SCcDKrjK0+Czfl1OysOta4nG/0\nOtyhhQRpVNdt+jj8VB9q8n0wUCMn8WcHj6P5M2adqoC+NT+F5htL+OxqspUGBY+QCsIpovrqFUtx\n7ZRT3wytb5q53xz9RmOliKpRvZCop5/oXPNj/f1ZYgeMrslWIqzhEVJBFKsNWt8MLSfmxQ9dnOfE\n9FNX1O3qNImT4SWs6Sc61/wUmm+A0TXZSkSS9AfQ1tamtm7dGvVtEFK22Lk466rrfBlmdF4rDAqb\n2YG0ESbMVgQdLs3CPsrCGl5upKfTFRoVIrJNKdXm5r1MaRJCsrgZORbFtcIgDtNPdOx9FBGMq0vl\niZtdTdZyhVpGGcsVat1HOULBI4Rk0enEjHIZrF/iMv0kaOS15OoZeW5Mu5qsLldokqDgEUKy6HRi\nmnB1OqFre3gc0BV5FRpUCl/rcoUmCZpWCCFZ3C5wdWNGCWsZrO7t4VFTLPLSiZP7048rNClQ8Agh\nWdw4MS0zSv+Rfiio7IixQtEzNS2mkHLbHh5W5KXTFZoUmNIkhORRqi/OixkljB67qPvndKOzH68Y\nVno06S5NL1DwCKlwvO6ji5sZRccg6TjVAJdPbc6r4QHmIi8drtAkwZQmIRWM2/RkLn4HPpsi6CDp\nuNUAb2w6E3efdw4m16YgACbXpnD3eedUlDCZgoJHSAXjZwNCWGYUt4ydPQGNN0z3vbcujjXAG5vO\nxNYrLkD/VR/G1isuoNhpgilNQioYP+nJIAOfTRGkf850DTBO6dJcymHKilcoeIRUMH575eIy8FkH\nppbJAqPHlVnpUgCRil4lTlkBmNIkpKKJW3oyCkwsk7WIY7oUCK/XL24wwiOkgoljejJsTM7QjGvL\nRCVOWQEoeIRUPOWUnvSLqRmaJtOlQQir1y9uMKVJCCGGMJkuDYKJKSuPDbyLtudeRfOzL6HtuVfx\n2MC7QW9TO4zwCCHEEHFYOWSH7ikrSTHBcAEsIYSUGWG3HLQ996ptinRybQpbr7jA2OcC3hbAuk5p\niki1iGwXkQ2Z11NE5AUR2S0i/yoiYxzOaxWR50XkVRH5jYjUZY5/MvO6V0T+TUTOdnsvhBBC7LGi\nrb3DI1A4FW2ZTDEmxQTjpYa3GMBrOa+/DaBbKTUdwO8BfKbwBBGpAfDPAD6rlLoAwJUARjLH7wVw\nlVKqFUAvgM/7egJCCCFZomg5SMqqIVeCJyKTAcwH8MPMawEwF8Cjmbc8COB6m1PnAehVSr0MAEqp\nd5RSJwBI5r+xmWuNA7A/wHMQQghBNNFWUlYNuY3w7gHwZQBWB+VZAA4ppY5nXu8FMMnmvBkAlIg8\nJSIvisiXAUApNQLgrwH8BmmhmwXgAbsPFpHbRWSriGw9ePCgy9slhJDKJIpoKykDr0sKnogsAHBA\nKbUt97DNW+3cLzUAPgrg1syvHxeRDhFJIS14swFMRDqludzu85VS9yul2pRSbePHjy91u4QQUtFE\nFW0lYeC1m7aEdgDXici1AOqQTj/eA6BRRGoyUd5k2Kck9wL4hVLqbQAQkScBXALgMAAopd7IHH8Y\nwLKAz0IIIRVPJS52dUtJwVNKLUcm+hKRKwEsVUrdKiKPALgJwFoAtwFYZ3P6UwC+LCKnATgG4E8B\ndAPYB2CWiIxXSh0EcDXyDTGEEEJ8UmmLXd0SZNLKnQC+KCK/Rbqm9wAAiMh1IrISAJRSvwfwvwD8\nGsBLAF5USm1USu0H8DUA/yEivQA+DOBbAe6FEEIIKQobzwkhhCQWI43nhBBCSJKh4BFCCKkIKHiE\nEEIqAgoeIYSQioCCRwghpCKg4BFCCKkIKHiEEEIqAgoeIYSQioCCRwghpCKg4BFCCKkIKHiEEEIq\nAgoeIYSQioCCRwghpCKg4BFCCKkIKHiEEEIqAgoeIYSQioCCRwghpCJI1MZzETkI4P+WeNvZAN4O\n4XZMUy7PAZTPs/A54ke5PEu5PAcQ/rN8SCk13s0bEyV4bhCRrW7XvceZcnkOoHyehc8RP8rlWcrl\nOYB4PwtTmoQQQioCCh4hhJCKoBwF7/6ob0AT5fIcQPk8C58jfpTLs5TLcwAxfpayq+ERQgghdpRj\nhEcIIYSMIhGCJyL/KiIvZf7rE5GXCr7+QRF5T0SWOpw/RUReEJHdmWuNyRz/oojsEJFeEdkkIh9K\n8LN8TEReFJHjInJTgp+jNvP6t5mvt0T1LCJyac7xl0Xk4w7nz8382b8iIg+KSE3m+AdE5CeZf1+/\nEpELE/ocDSKyPnPuqyKyyORzGH6WO3LOf0VETojImUl7jszXrsyc/6qI/MLUM5h8jswzDOZco8vk\nc0Aplaj/APwdgK6CY48BeATAUodzHgZwS+b3PwDw15nfXwXgtMzv/xrAvyb4WVoAtAJ4CMBNCX6O\nvwHwg8zvb4ny7wTAaQBqMr9vBnDAep3z/ioAbwGYkXm9EsBnMr//DoD/mfn9TACbEvocdwH4dub3\n4wG8C2BMEp+l4H2dADYn8TkANALYAeCDmdcTEvocVwLYENa9JyLCsxARAXAzgB/nHLsewB4ArxY5\nZy6ARzOHHgRwPQAopZ5VSr2fOf5LAJPN3Lnjfel8lj6lVC+AkwZv2+metD0HgIWZ18h8vSPzfuMU\nPotS6n2l1PHMl+sA2BW8zwIwrJTalXn9DIAbM7+fBWBT5lo7AbSIyB8Zuv0sBp5DATgjc93TkRa8\n46MvoR8Dz5LLJ5Hz79YkBp7jUwAeV0r9V+Z6B0zdey6G/z6MkyjBAzAHwO+UUrsBQETGArgTwNeK\nnHMWgEM5fyl7AUyyed9nAPxM472WwuSzhInu55iE9E+DyHx9MPP+MMh7FgAQkctE5FUAvwHw2Zx7\ntngbQEpErEbbmwCck/n9ywBuyFznUgAfQjg/VOl+ju8COB/A/sz5i5VSYf1gpftZrGucBuC/IZ2J\nCAPdzzEDwAdE5Ocisk1EPm34/i1M/H1cnkmH/kxELjB58zWl3xIOIvLvAJpsvvQVpdS6zO8LfyL7\nGoBupdR7RYIAuy/k/RQiIv8dQBuAP/V0004fGOGz6CSi5zDyjD6fBUqpFwBcICLnA3hQRH6mlBrK\n+boSkVsAdItILYCncSr6WQ3g3ky94zcAtiNgZBTRc1wD4CWko/JpAJ4RkS1KqcMJfBaLTgA9Sql3\ngzxDhM9RA+CPAXQAqAfwvIj8MieKSspzvIj0aLD3RORaAE8AmO73GUoSVu406H9I/wX/DsDknGNb\nAPRl/juEdKrl8wXnCdI/YVh55ssBPJXz9T8D8BrCzYEbeZbMsX9ESDU8E88B4CkAl+dc/21k2mfC\nfhab9zwLoK3EdeYBeNjmuGT+TMYl7TkAbAQwJ+drmwFcmuS/EwA/AfAp089g8O9kGYAVOV97AMAn\nkvYcNl/rA3C2sWcI4y9c0x/2fwPwiyJfXwFng8QjyDdI/E3m97MBvAFgetKfJefr/4jwBM/E38nn\nkG9asf0fI4xnATAFp0T5Q0in9Eb9z4jMD0sAapGu2c3NvG5ExtwB4K8APJTQ5/i+9c0VwB8B2Gfy\nm5LJZ8kca0D6B7GxCf63dX7mdQ3SxpFXAFyYwOdowql+8EsB/BcM/oCbpBreLfBQYBaRJ0VkYubl\nnQC+KCK/Rboe9EDm+HeQLsI/krHE/lTnDRdB+7OIyJ+IyF4AnwDwvzM5ddOY+Dt5AMBZmeNfRPon\n2TCwe5aPAng5k5L8CdKi/DYw6lnuEJHXAPQCWK+U2pw5fj6AV0VkJ4A/B7DY9EPAzHN8HcAVIvIb\npL9Z3WmdbxgTzwIAHwfwtFLqiNnbz6L9OZRSrwH4t8zxXwH4oVLqlaQ9B9L1vFdE5GUA9yH9Q7C5\nMo3BaxNCCCGxIUkRHiGEEOIbCh4hhJCKgIJHCCGkIqDgEUIIqQgoeIQQQioCCh4hhJCKgIJHCCGk\nIqDgEUIIqQj+f0yhlqnjcDlmAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x16856297710>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "K1, K2, K = 24, 8, 10\n",
    "X = np.array(station[['latitude', 'longitude']])\n",
    "kmeans = KMeans(n_clusters=K1).fit(X)\n",
    "station['bcluster'] = kmeans.labels_\n",
    "bc_df = nyc_df[['starttime','start station id','end station id']]\n",
    "bc_df = bc_df.merge(station[['id', 'bcluster']],\n",
    "            how='left',left_on='end station id',right_on='id').drop('id',axis=1)\n",
    "plt.figure(figsize=(7,7))\n",
    "_centers = kmeans.cluster_centers_.T\n",
    "plt.scatter(_centers[1], _centers[0], marker='x')\n",
    "_cluster = list(station.groupby('bcluster').groups.values())\n",
    "for c in _cluster:\n",
    "    plt.scatter(station.loc[c]['longitude'], station.loc[c]['latitude'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "#### Transit-matrix generation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Maximum likelyhood estimation**\n",
    "\n",
    "the pmf of **multinormial distribution**: \n",
    "$$f(\\boldsymbol{X},N;\\boldsymbol{\\theta})=\n",
    "\\binom{N}{x_1,x_2,\\cdots,x_n}\n",
    "\\theta_1^{x_1}\\theta_2^{x_2}\\cdots\\theta_n^{x_n}\n",
    "$$\n",
    "here, $\n",
    "\\boldsymbol{X}=(x_1,x_2,\\cdots,x_n),\n",
    "\\boldsymbol{\\theta}=(\\theta_1,\\theta_2,\\cdots,\\theta_n),\n",
    "\\sum\\theta_j=1\n",
    "$\n",
    "\n",
    "for each sample $\\boldsymbol{X}^i$ checks in at Cluster $C_k$, \n",
    "$$N=1\\quad x_k^i=1\\qquad x_j^i=0, \\forall j \\ne k$$\n",
    "\n",
    "suppose the number of the samples that check in at Clusters\n",
    "$(C_1,C_2,\\cdots,C_m)$ is $(l_1, l_2, \\cdots, l_m)$ respectively.\n",
    "\n",
    "the likelyhood function:\n",
    "$$L(\\boldsymbol{\\theta})=\n",
    "\\prod_i^{T}p(\\boldsymbol{x}^i,1;\\boldsymbol{\\theta})=\\prod_{j=1}^m\\theta_j^{l_j}$$\n",
    "$$\\ln L(\\boldsymbol{\\theta})\n",
    "%=\\sum_i^{T}\\ln p(\\boldsymbol{X}_i;\\boldsymbol{\\theta})\n",
    "=\\sum_j^ml_j\\ln\\theta_j\n",
    "$$\n",
    "Now maximize \n",
    "$\\quad\\ln L(\\boldsymbol{\\theta})\n",
    "\\quad\\mathrm{s.t.}\\ \\sum\\theta_j=1\n",
    "$\n",
    "\n",
    "by applying **Lagrange multiplier method**\n",
    "\n",
    "let $F(\\boldsymbol{\\theta},\\lambda)=\\ln L(\\boldsymbol{\\theta})-\\lambda g(\\boldsymbol{\\theta}),\\quad\n",
    "g(\\boldsymbol{\\theta})=\\sum\\theta_j-1$\n",
    "\n",
    "$$\\begin{aligned}\n",
    "&\\frac{\\partial F(\\boldsymbol{\\theta},\\lambda)}{\\partial\\theta_j}=\\frac{l_j}{\\theta_j}-\\lambda=0\\\\\n",
    "&\\theta_j=\\lambda^{-1}l_j\\\\\n",
    "\\therefore\\quad&\\theta_j\\propto l_j\\\\\n",
    "&\\theta_k=\\frac{l_k}{\\sum l_j}\n",
    "\\end{aligned}$$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def calc_slot(raw_date_str):\n",
    "    dt = datetime.strptime(raw_date_str, '%m/%d/%Y %H:%M')\n",
    "    weekday = dt.weekday()\n",
    "    hr = dt.hour\n",
    "    if weekday < 5:\n",
    "        if hr >= 7:\n",
    "            if hr < 11:\n",
    "                return 0\n",
    "            if hr < 16:\n",
    "                return 1\n",
    "            if hr < 21:\n",
    "                return 2\n",
    "        return 3\n",
    "    if hr < 9:\n",
    "         return 4\n",
    "    if hr < 19:\n",
    "        return 5\n",
    "    return 6\n",
    "def get_t_vector(station_id):\n",
    "    st = bc_df[bc_df['start station id'] == station_id][:]\n",
    "    st['slot'] =  st['starttime'].map(calc_slot)\n",
    "    #st = st.drop('starttime', axis=1)\n",
    "    by_slots = dict(st.groupby('slot').groups)\n",
    "    for key in by_slots.keys():\n",
    "        ndict = dict(st.loc[by_slots[key]].groupby('bcluster').size())\n",
    "        count = np.array(list(map(lambda x:ndict[x] if x in ndict else 0, range(K1))))\n",
    "        count = count / count.sum()\n",
    "        by_slots[key] = count\n",
    "    if len(by_slots) < 7:\n",
    "        mean = np.array(list(by_slots.values())).mean(axis=0)\n",
    "        for i in range(7):\n",
    "            if i not in by_slots.keys():\n",
    "                by_slots[i] = mean\n",
    "    return np.array(list(by_slots.values())).reshape(7*K1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 0.1308642 ,  0.06419753,  0.        ,  0.01975309,  0.        ,\n",
       "         0.12592593,  0.05679012,  0.01728395,  0.        ,  0.        ,\n",
       "         0.04444444,  0.14320988,  0.09876543,  0.00987654,  0.        ,\n",
       "         0.        ,  0.        ,  0.03950617,  0.        ,  0.08888889,\n",
       "         0.05679012,  0.        ,  0.1037037 ,  0.        ],\n",
       "       [ 0.22406639,  0.0746888 ,  0.        ,  0.01659751,  0.        ,\n",
       "         0.1659751 ,  0.01659751,  0.00414938,  0.        ,  0.00414938,\n",
       "         0.00829876,  0.1659751 ,  0.05394191,  0.00414938,  0.        ,\n",
       "         0.        ,  0.00414938,  0.04149378,  0.        ,  0.09958506,\n",
       "         0.0373444 ,  0.        ,  0.0746888 ,  0.00414938],\n",
       "       [ 0.1475    ,  0.08      ,  0.        ,  0.0375    ,  0.0075    ,\n",
       "         0.1075    ,  0.0575    ,  0.005     ,  0.        ,  0.0025    ,\n",
       "         0.0325    ,  0.165     ,  0.0525    ,  0.0375    ,  0.        ,\n",
       "         0.        ,  0.0075    ,  0.0375    ,  0.        ,  0.125     ,\n",
       "         0.055     ,  0.        ,  0.035     ,  0.0075    ],\n",
       "       [ 0.29824561,  0.01754386,  0.        ,  0.        ,  0.        ,\n",
       "         0.10526316,  0.06140351,  0.02631579,  0.        ,  0.        ,\n",
       "         0.07894737,  0.15789474,  0.        ,  0.01754386,  0.        ,\n",
       "         0.        ,  0.00877193,  0.05263158,  0.        ,  0.07894737,\n",
       "         0.07894737,  0.        ,  0.00877193,  0.00877193],\n",
       "       [ 0.28571429,  0.07142857,  0.        ,  0.        ,  0.        ,\n",
       "         0.03571429,  0.        ,  0.07142857,  0.        ,  0.        ,\n",
       "         0.07142857,  0.10714286,  0.        ,  0.07142857,  0.        ,\n",
       "         0.03571429,  0.        ,  0.03571429,  0.        ,  0.14285714,\n",
       "         0.        ,  0.        ,  0.07142857,  0.        ],\n",
       "       [ 0.19504644,  0.11455108,  0.        ,  0.01547988,  0.        ,\n",
       "         0.17337461,  0.04643963,  0.00619195,  0.        ,  0.00619195,\n",
       "         0.03095975,  0.10216718,  0.0247678 ,  0.0247678 ,  0.00309598,\n",
       "         0.        ,  0.00619195,  0.03405573,  0.        ,  0.05572755,\n",
       "         0.08049536,  0.        ,  0.06501548,  0.01547988],\n",
       "       [ 0.18309859,  0.04225352,  0.        ,  0.04225352,  0.        ,\n",
       "         0.05633803,  0.05633803,  0.        ,  0.        ,  0.        ,\n",
       "         0.08450704,  0.09859155,  0.05633803,  0.04225352,  0.        ,\n",
       "         0.        ,  0.04225352,  0.01408451,  0.        ,  0.04225352,\n",
       "         0.15492958,  0.        ,  0.08450704,  0.        ]])"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "get_t_vector(72).reshape(7,K1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAbwAAAGfCAYAAAAkiGdHAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngIJHCCGkElDwCCGEVIL/D8qjd/RCTxP/AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1685782c978>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "X = np.array(list(station['id'].map(get_t_vector)))\n",
    "station['bcluster'] = KMeans(n_clusters=K2).fit(X).labels_\n",
    "\n",
    "plt.figure(figsize=(7,7))\n",
    "_cluster = list(station.groupby('bcluster').groups.values())\n",
    "for c in _cluster:\n",
    "    plt.scatter(station.loc[c]['longitude'], station.loc[c]['latitude'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5,1,'BC (first iteration, not convergent yet)')"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAbwAAAGrCAYAAAC7YyL8AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nTYEKnqIoitIUqOApiqIoTcH/BeLOJYsbre85AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x168577a2cf8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "n = len(station)\n",
    "vals = station.groupby('bcluster').groups.values()\n",
    "offset = 0\n",
    "for v in vals:\n",
    "    X = np.array(station.loc[v][['latitude', 'longitude']])\n",
    "    n_c = int(np.round(len(v)*K1/n))\n",
    "    station.loc[v,'bcluster'] = KMeans(n_clusters=n_c).fit(X).labels_ + offset\n",
    "    offset += n_c\n",
    "\n",
    "plt.figure(figsize=(7,7))\n",
    "_cluster = list(station.groupby('bcluster').groups.values())\n",
    "for c in _cluster:\n",
    "    plt.scatter(station.loc[c]['longitude'], station.loc[c]['latitude'])\n",
    "plt.title('BC (first iteration, not convergent yet)')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "def sameCluster(s, s0):\n",
    "    A = sorted(s.groupby('bcluster').groups.values(), key = lambda x: x[0])\n",
    "    B = sorted(s0.groupby('bcluster').groups.values(), key = lambda x: x[0])\n",
    "    if len(A) != len(B):\n",
    "        return False\n",
    "    for a, b in zip(A, B):\n",
    "        if len(a) != len(b):\n",
    "            return False\n",
    "        if (a != b).sum() != 0:\n",
    "            return False\n",
    "    return True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5,1,'BC')"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAbwAAAGrCAYAAAC7YyL8AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nDBsfOFLTAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x168466953c8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "n = len(station)\n",
    "K1, K2, K = 24, 8, 5\n",
    "X = np.array(station[['latitude', 'longitude']])\n",
    "station['bcluster'] = KMeans(n_clusters=K1).fit(X).labels_\n",
    "for k in range(K):\n",
    "    bc_df = nyc_df[['starttime','start station id','end station id']]\n",
    "    bc_df = bc_df.merge(station[['id', 'bcluster']],\n",
    "                how='left',left_on='end station id',right_on='id').drop('id',axis=1)\n",
    "    last_station = station[['id', 'bcluster']]\n",
    "    X = np.array(list(station['id'].map(get_t_vector)))\n",
    "    station['bcluster'] = KMeans(n_clusters=K2).fit(X).labels_\n",
    "    vals = station.groupby('bcluster').groups.values()\n",
    "    offset = 0\n",
    "    for v in vals:\n",
    "        X = np.array(station.loc[v][['latitude', 'longitude']])\n",
    "        n_c = int(np.round(len(v)*K1/n))\n",
    "        station.loc[v,'bcluster'] = KMeans(n_clusters=n_c).fit(X).labels_ + offset\n",
    "        offset += n_c\n",
    "    if sameCluster(station, last_station):\n",
    "        break\n",
    "K1 = offset # offset might be K1 + 1 because of rounding error\n",
    "plt.figure(figsize=(7,7))\n",
    "_cluster = list(station.groupby('bcluster').groups.values())\n",
    "for c in _cluster:\n",
    "    plt.scatter(station.loc[c]['longitude'], station.loc[c]['latitude'])\n",
    "plt.title('BC')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "    id                           name  gcluster  bcluster\n",
      "0   72               W 52 St & 11 Ave        22        20\n",
      "1   79       Franklin St & W Broadway         9        10\n",
      "2   82         St James Pl & Pearl St        10         9\n",
      "3   83  Atlantic Ave & Fort Greene Pl         1         1\n",
      "4  116                W 17 St & 8 Ave        15        18\n"
     ]
    }
   ],
   "source": [
    "cluster_df = station[['id', 'name', 'gcluster', 'bcluster']]\n",
    "print(cluster_df.head())\n",
    "cluster_df.to_csv(path+'clusters.csv',index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "## Entire Traffic Learning\n",
    "### Time features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0    8/1/2014 0:00\n",
      "1    8/1/2014 0:00\n",
      "2    8/1/2014 0:00\n",
      "3    8/1/2014 0:00\n",
      "4    8/1/2014 0:00\n",
      "Name: starttime, dtype: object\n",
      "473615    8/15/2014 23:59\n",
      "473616    8/15/2014 23:59\n",
      "473617    8/15/2014 23:59\n",
      "473618    8/15/2014 23:59\n",
      "473619    8/15/2014 23:59\n",
      "Name: starttime, dtype: object\n"
     ]
    }
   ],
   "source": [
    "print(nyc_df['starttime'].head())\n",
    "print(nyc_df['starttime'].tail())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "day  hour\n",
      "1    0        393\n",
      "     1        171\n",
      "     2        126\n",
      "     3         53\n",
      "     4         69\n",
      "     5        208\n",
      "     6        823\n",
      "     7       1632\n",
      "     8       2950\n",
      "     9       2272\n",
      "     10      1365\n",
      "     11      1406\n",
      "     12      1682\n",
      "     13      1884\n",
      "     14      1922\n",
      "     15      1952\n",
      "     16      2496\n",
      "     17      2958\n",
      "     18      2711\n",
      "     19      1966\n",
      "     20      1329\n",
      "     21       953\n",
      "     22       722\n",
      "     23       611\n",
      "2    0        481\n",
      "     1        346\n",
      "     2        201\n",
      "     3        101\n",
      "     4         29\n",
      "     5         17\n",
      "             ... \n",
      "14   18      3726\n",
      "     19      2598\n",
      "     20      1780\n",
      "     21      1192\n",
      "     22       933\n",
      "     23       745\n",
      "15   0        397\n",
      "     1        184\n",
      "     2         94\n",
      "     3         50\n",
      "     4         68\n",
      "     5        189\n",
      "     6        766\n",
      "     7       1526\n",
      "     8       2808\n",
      "     9       2248\n",
      "     10      1416\n",
      "     11      1381\n",
      "     12      1680\n",
      "     13      1823\n",
      "     14      1944\n",
      "     15      2168\n",
      "     16      2527\n",
      "     17      3189\n",
      "     18      2606\n",
      "     19      1900\n",
      "     20      1217\n",
      "     21       882\n",
      "     22       769\n",
      "     23       648\n",
      "Length: 360, dtype: int64\n"
     ]
    }
   ],
   "source": [
    "series = nyc_df['starttime'].map(lambda x:datetime.strptime(x, '%m/%d/%Y %H:%M'))\n",
    "time_df = pd.DataFrame({'starttime': series})\n",
    "time_df['day'] = time_df['starttime'].map(lambda x:x.day)\n",
    "time_df['hour'] = time_df['starttime'].map(lambda x:x.hour)\n",
    "by_hour = time_df.groupby(['day','hour'])\n",
    "print(by_hour.size())    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "     entiretraffic  year  month  day  hour  weekday\n",
      "0              393  2014      8    1     0        4\n",
      "393            171  2014      8    1     1        4\n",
      "564            126  2014      8    1     2        4\n",
      "690             53  2014      8    1     3        4\n",
      "743             69  2014      8    1     4        4\n"
     ]
    }
   ],
   "source": [
    "vals = by_hour.groups.values()\n",
    "first_id = list(map(lambda x:x[0],vals))\n",
    "create = {}\n",
    "create['entiretraffic'] = list(by_hour.size())\n",
    "create['datetime']= time_df.loc[first_id, 'starttime']\n",
    "train_df = pd.DataFrame(create)\n",
    "train_df['year'] = train_df['datetime'].map(lambda x:x.year)\n",
    "train_df['month'] = train_df['datetime'].map(lambda x:x.month)\n",
    "train_df['day'] = train_df['datetime'].map(lambda x:x.day)\n",
    "train_df['hour'] = train_df['datetime'].map(lambda x:x.hour)\n",
    "train_df['weekday'] = train_df['datetime'].map(lambda x:x.weekday())\n",
    "train_df = train_df.drop('datetime', axis=1)\n",
    "print(train_df.head())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Meteorology features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>time</th>\n",
       "      <th>Temperature ° F</th>\n",
       "      <th>Wind Speed mph</th>\n",
       "      <th>Weather</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>8-1-2014 0:51 EDT</td>\n",
       "      <td>73.9</td>\n",
       "      <td>3.5</td>\n",
       "      <td>no value</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>8-1-2014 1:51 EDT</td>\n",
       "      <td>73.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>no value</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>8-1-2014 2:51 EDT</td>\n",
       "      <td>72.0</td>\n",
       "      <td>3.5</td>\n",
       "      <td>no value</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>8-1-2014 3:51 EDT</td>\n",
       "      <td>72.0</td>\n",
       "      <td>3.5</td>\n",
       "      <td>no value</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>8-1-2014 4:51 EDT</td>\n",
       "      <td>71.1</td>\n",
       "      <td>4.6</td>\n",
       "      <td>no value</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                time  Temperature ° F  Wind Speed mph   Weather\n",
       "0  8-1-2014 0:51 EDT             73.9             3.5  no value\n",
       "1  8-1-2014 1:51 EDT             73.0             0.0  no value\n",
       "2  8-1-2014 2:51 EDT             72.0             3.5  no value\n",
       "3  8-1-2014 3:51 EDT             72.0             3.5  no value\n",
       "4  8-1-2014 4:51 EDT             71.1             4.6  no value"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mete_df = pd.read_excel(path+'Meteorology-NYC.xls')\n",
    "mete_df = mete_df.reset_index().rename(columns={'index':'time'})\n",
    "mete_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Temperature ° F</th>\n",
       "      <th>Wind Speed mph</th>\n",
       "      <th>Weather</th>\n",
       "      <th>year</th>\n",
       "      <th>month</th>\n",
       "      <th>day</th>\n",
       "      <th>hour</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>73.9</td>\n",
       "      <td>3.5</td>\n",
       "      <td>no value</td>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>73.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>no value</td>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>72.0</td>\n",
       "      <td>3.5</td>\n",
       "      <td>no value</td>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>72.0</td>\n",
       "      <td>3.5</td>\n",
       "      <td>no value</td>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>71.1</td>\n",
       "      <td>4.6</td>\n",
       "      <td>no value</td>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Temperature ° F  Wind Speed mph   Weather  year  month  day  hour\n",
       "0             73.9             3.5  no value  2014      8    1     0\n",
       "1             73.0             0.0  no value  2014      8    1     1\n",
       "2             72.0             3.5  no value  2014      8    1     2\n",
       "3             72.0             3.5  no value  2014      8    1     3\n",
       "4             71.1             4.6  no value  2014      8    1     4"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mete_df['datetime'] = mete_df['time'].map(lambda x:datetime.strptime(x.strip(' EDT'), '%m-%d-%Y %H:%M'))\n",
    "mete_df['year'] = mete_df['datetime'].map(lambda x:x.year)\n",
    "mete_df['month'] = mete_df['datetime'].map(lambda x:x.month)\n",
    "mete_df['day'] = mete_df['datetime'].map(lambda x:x.day)\n",
    "mete_df['hour'] = mete_df['datetime'].map(lambda x:x.hour)\n",
    "mete_df = mete_df.drop(['time', 'datetime'], axis=1)\n",
    "mete_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Temperature ° F     0\n",
       "Wind Speed mph     28\n",
       "Weather             0\n",
       "year                0\n",
       "month               0\n",
       "day                 0\n",
       "hour                0\n",
       "dtype: int64"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# check if there's missing data\n",
    "mete_df.isnull().sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Wind Speed mph</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>444</th>\n",
       "      <td>9.2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>445</th>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>446</th>\n",
       "      <td>4.6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>447</th>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>448</th>\n",
       "      <td>3.5</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     Wind Speed mph\n",
       "444             9.2\n",
       "445             NaN\n",
       "446             4.6\n",
       "447             NaN\n",
       "448             3.5"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "wind = mete_df[['Wind Speed mph']][:]\n",
    "wind[wind == 0] = float('nan')\n",
    "wind.tail()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Wind Speed mph</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>444</th>\n",
       "      <td>9.20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>445</th>\n",
       "      <td>6.90</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>446</th>\n",
       "      <td>4.60</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>447</th>\n",
       "      <td>4.05</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>448</th>\n",
       "      <td>3.50</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     Wind Speed mph\n",
       "444            9.20\n",
       "445            6.90\n",
       "446            4.60\n",
       "447            4.05\n",
       "448            3.50"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "forward = wind.fillna(method='ffill')\n",
    "backward = wind.fillna(method='bfill')\n",
    "wind = ((forward + backward) / 2)\n",
    "mete_df['Wind Speed mph'] = wind\n",
    "wind.tail()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['no value', 'lt rain; fog', 'fog', 'mod rain; fog', 'lt rain', 'haze'], dtype='object')"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# check how many weather type\n",
    "mete_df['Weather'].value_counts().index"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "mete_df['weather'] = -1\n",
    "mete_df.loc[mete_df.Weather == 'no value', 'weather'] = 3\n",
    "mete_df.loc[mete_df.Weather == 'haze', 'weather'] = 2\n",
    "mete_df.loc[mete_df.Weather == 'fog', 'weather'] = 2\n",
    "mete_df.loc[mete_df.Weather == 'lt rain', 'weather'] = 1\n",
    "mete_df.loc[mete_df.Weather == 'lt rain; fog', 'weather'] = 1\n",
    "mete_df.loc[mete_df.Weather == 'mod rain; fog', 'weather'] = 1\n",
    "mete_df = mete_df.drop('Weather', axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Note that len(mete_df) = 449 > 15*24=360\n",
    "# This is because in some hours, there are more one entry.\n",
    "# e.g. 8-2-2014 5:26, EDT 8-2-2014 5:49, EDT 8-2-2014 5:51 EDT\n",
    "# Remove duplicate and replace by average data\n",
    "tmp = mete_df.groupby(['year', 'month', 'day', 'hour']).mean().reset_index()\n",
    "tmp['weather'] = mete_df.groupby(['year', 'month', 'day', 'hour']).max().reset_index()['weather']\n",
    "mete_df = tmp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Int64Index([3, 1, 2], dtype='int64')"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mete_df['weather'].value_counts().index"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### GBRT training and prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>entiretraffic</th>\n",
       "      <th>year</th>\n",
       "      <th>month</th>\n",
       "      <th>day</th>\n",
       "      <th>hour</th>\n",
       "      <th>weekday</th>\n",
       "      <th>Temperature ° F</th>\n",
       "      <th>Wind Speed mph</th>\n",
       "      <th>weather</th>\n",
       "      <th>isweekend</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>393</td>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>73.9</td>\n",
       "      <td>3.5</td>\n",
       "      <td>3</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>171</td>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>73.0</td>\n",
       "      <td>3.5</td>\n",
       "      <td>3</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>126</td>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>4</td>\n",
       "      <td>72.0</td>\n",
       "      <td>3.5</td>\n",
       "      <td>3</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>53</td>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>4</td>\n",
       "      <td>72.0</td>\n",
       "      <td>3.5</td>\n",
       "      <td>3</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>69</td>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>4</td>\n",
       "      <td>71.1</td>\n",
       "      <td>4.6</td>\n",
       "      <td>3</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   entiretraffic  year  month  day  hour  weekday  Temperature ° F  \\\n",
       "0            393  2014      8    1     0        4             73.9   \n",
       "1            171  2014      8    1     1        4             73.0   \n",
       "2            126  2014      8    1     2        4             72.0   \n",
       "3             53  2014      8    1     3        4             72.0   \n",
       "4             69  2014      8    1     4        4             71.1   \n",
       "\n",
       "   Wind Speed mph  weather  isweekend  \n",
       "0             3.5        3      False  \n",
       "1             3.5        3      False  \n",
       "2             3.5        3      False  \n",
       "3             3.5        3      False  \n",
       "4             4.6        3      False  "
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df = train_df.merge(mete_df, how='left',on=['year','month','day','hour'])\n",
    "train_df['isweekend'] = train_df['weekday'].map(lambda x:x>4)\n",
    "train_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "360 360\n"
     ]
    }
   ],
   "source": [
    "print(len(train_df), len(mete_df))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "train = train_df.head(320)\n",
    "train_X = train.drop('entiretraffic',axis=1)\n",
    "train_y = train['entiretraffic']\n",
    "test = train_df.tail(40)\n",
    "test_X = test.drop('entiretraffic',axis=1)\n",
    "test_y = test['entiretraffic']\n",
    "\n",
    "dtrain = xgb.DMatrix(train_X, train_y)\n",
    "dtest = xgb.DMatrix(test_X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "xgb_params = {\n",
    "    'eta': 0.01,\n",
    "    'max_depth': 7,\n",
    "    'subsample': 1.0,\n",
    "    'colsample_bytree': 1.0,\n",
    "    'objective': 'reg:linear',\n",
    "    'eval_metric': 'rmse',\n",
    "    'silent': 1\n",
    "}\n",
    "model = xgb.train(xgb_params, dtrain, num_boost_round=5000)\n",
    "pred_y = model.predict(dtest)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYAAAAD8CAYAAAB+UHOxAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x16857845908>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(range(len(test_y)),test_y)\n",
    "plt.plot(range(len(test_y)),pred_y)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Cluster Check-out Proportion Learning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "tmp_df = nyc_df[['starttime','start station id']]\n",
    "tmp_df = tmp_df.merge(station[['id', 'bcluster']],\n",
    "            how='left',left_on='start station id',right_on='id').drop('id',axis=1)\n",
    "tmp_df['sdt'] = tmp_df['starttime'].map(lambda x:datetime.strptime(x, '%m/%d/%Y %H:%M'))\n",
    "tmp_df['year'] = tmp_df['sdt'].map(lambda x:x.year)\n",
    "tmp_df['month'] = tmp_df['sdt'].map(lambda x:x.month)\n",
    "tmp_df['day'] = tmp_df['sdt'].map(lambda x:x.day)\n",
    "tmp_df['hour'] = tmp_df['sdt'].map(lambda x:x.hour)\n",
    "\n",
    "by_ymdh = dict(tmp_df.groupby(['year','month','day','hour']).groups)\n",
    "cnt_dict = {}\n",
    "for key in by_ymdh.keys():\n",
    "    ndict = dict(tmp_df.loc[by_ymdh[key]].groupby('bcluster').size())\n",
    "    count = np.array(list(map(lambda x:ndict[x] if x in ndict else 0, range(K1))))\n",
    "    cnt_dict[key] = count\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>level_0</th>\n",
       "      <th>level_1</th>\n",
       "      <th>level_2</th>\n",
       "      <th>level_3</th>\n",
       "      <th>weekday</th>\n",
       "      <th>isweekend</th>\n",
       "      <th>proportion</th>\n",
       "      <th>entiretraffic</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>False</td>\n",
       "      <td>[0.0, 0.0178117048346, 0.0559796437659, 0.0839...</td>\n",
       "      <td>393</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>False</td>\n",
       "      <td>[0.00584795321637, 0.0116959064327, 0.07017543...</td>\n",
       "      <td>171</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>4</td>\n",
       "      <td>False</td>\n",
       "      <td>[0.047619047619, 0.00793650793651, 0.071428571...</td>\n",
       "      <td>126</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>4</td>\n",
       "      <td>False</td>\n",
       "      <td>[0.0, 0.0188679245283, 0.0377358490566, 0.1132...</td>\n",
       "      <td>53</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>4</td>\n",
       "      <td>False</td>\n",
       "      <td>[0.0144927536232, 0.0144927536232, 0.028985507...</td>\n",
       "      <td>69</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   level_0  level_1  level_2  level_3  weekday  isweekend  \\\n",
       "0     2014        8        1        0        4      False   \n",
       "1     2014        8        1        1        4      False   \n",
       "2     2014        8        1        2        4      False   \n",
       "3     2014        8        1        3        4      False   \n",
       "4     2014        8        1        4        4      False   \n",
       "\n",
       "                                          proportion  entiretraffic  \n",
       "0  [0.0, 0.0178117048346, 0.0559796437659, 0.0839...            393  \n",
       "1  [0.00584795321637, 0.0116959064327, 0.07017543...            171  \n",
       "2  [0.047619047619, 0.00793650793651, 0.071428571...            126  \n",
       "3  [0.0, 0.0188679245283, 0.0377358490566, 0.1132...             53  \n",
       "4  [0.0144927536232, 0.0144927536232, 0.028985507...             69  "
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "series = pd.Series(cnt_dict)\n",
    "start_out_df = pd.DataFrame({'count': series}).reset_index()\n",
    "first_id = list(map(lambda x:x[0], by_ymdh.values()))\n",
    "start_out_df['weekday'] = list(tmp_df.loc[first_id, 'sdt'].map(lambda x:x.weekday()))\n",
    "start_out_df['isweekend'] = start_out_df['weekday'].map(lambda x:x>4)\n",
    "start_out_df['proportion'] = start_out_df['count'].map(lambda x:x/x.sum())\n",
    "start_out_df['entiretraffic'] = start_out_df['count'].map(lambda x:x.sum())\n",
    "start_out_df = start_out_df.drop('count', axis=1)\n",
    "\n",
    "start_out_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>year</th>\n",
       "      <th>month</th>\n",
       "      <th>day</th>\n",
       "      <th>hour</th>\n",
       "      <th>weekday</th>\n",
       "      <th>isweekend</th>\n",
       "      <th>proportion</th>\n",
       "      <th>entiretraffic</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>False</td>\n",
       "      <td>[0.0, 0.0178117048346, 0.0559796437659, 0.0839...</td>\n",
       "      <td>393</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>False</td>\n",
       "      <td>[0.00584795321637, 0.0116959064327, 0.07017543...</td>\n",
       "      <td>171</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>4</td>\n",
       "      <td>False</td>\n",
       "      <td>[0.047619047619, 0.00793650793651, 0.071428571...</td>\n",
       "      <td>126</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>4</td>\n",
       "      <td>False</td>\n",
       "      <td>[0.0, 0.0188679245283, 0.0377358490566, 0.1132...</td>\n",
       "      <td>53</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>4</td>\n",
       "      <td>False</td>\n",
       "      <td>[0.0144927536232, 0.0144927536232, 0.028985507...</td>\n",
       "      <td>69</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   year  month  day  hour  weekday  isweekend  \\\n",
       "0  2014      8    1     0        4      False   \n",
       "1  2014      8    1     1        4      False   \n",
       "2  2014      8    1     2        4      False   \n",
       "3  2014      8    1     3        4      False   \n",
       "4  2014      8    1     4        4      False   \n",
       "\n",
       "                                          proportion  entiretraffic  \n",
       "0  [0.0, 0.0178117048346, 0.0559796437659, 0.0839...            393  \n",
       "1  [0.00584795321637, 0.0116959064327, 0.07017543...            171  \n",
       "2  [0.047619047619, 0.00793650793651, 0.071428571...            126  \n",
       "3  [0.0, 0.0188679245283, 0.0377358490566, 0.1132...             53  \n",
       "4  [0.0144927536232, 0.0144927536232, 0.028985507...             69  "
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "start_out_df = start_out_df.rename(columns={'level_0':'year', 'level_1':'month', 'level_2':'day', 'level_3':'hour'})\n",
    "start_out_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>year</th>\n",
       "      <th>month</th>\n",
       "      <th>day</th>\n",
       "      <th>hour</th>\n",
       "      <th>weekday</th>\n",
       "      <th>isweekend</th>\n",
       "      <th>proportion</th>\n",
       "      <th>entiretraffic</th>\n",
       "      <th>Temperature ° F</th>\n",
       "      <th>Wind Speed mph</th>\n",
       "      <th>weather</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>False</td>\n",
       "      <td>[0.0, 0.0178117048346, 0.0559796437659, 0.0839...</td>\n",
       "      <td>393</td>\n",
       "      <td>73.9</td>\n",
       "      <td>3.5</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>False</td>\n",
       "      <td>[0.00584795321637, 0.0116959064327, 0.07017543...</td>\n",
       "      <td>171</td>\n",
       "      <td>73.0</td>\n",
       "      <td>3.5</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>4</td>\n",
       "      <td>False</td>\n",
       "      <td>[0.047619047619, 0.00793650793651, 0.071428571...</td>\n",
       "      <td>126</td>\n",
       "      <td>72.0</td>\n",
       "      <td>3.5</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>4</td>\n",
       "      <td>False</td>\n",
       "      <td>[0.0, 0.0188679245283, 0.0377358490566, 0.1132...</td>\n",
       "      <td>53</td>\n",
       "      <td>72.0</td>\n",
       "      <td>3.5</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>4</td>\n",
       "      <td>False</td>\n",
       "      <td>[0.0144927536232, 0.0144927536232, 0.028985507...</td>\n",
       "      <td>69</td>\n",
       "      <td>71.1</td>\n",
       "      <td>4.6</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   year  month  day  hour  weekday  isweekend  \\\n",
       "0  2014      8    1     0        4      False   \n",
       "1  2014      8    1     1        4      False   \n",
       "2  2014      8    1     2        4      False   \n",
       "3  2014      8    1     3        4      False   \n",
       "4  2014      8    1     4        4      False   \n",
       "\n",
       "                                          proportion  entiretraffic  \\\n",
       "0  [0.0, 0.0178117048346, 0.0559796437659, 0.0839...            393   \n",
       "1  [0.00584795321637, 0.0116959064327, 0.07017543...            171   \n",
       "2  [0.047619047619, 0.00793650793651, 0.071428571...            126   \n",
       "3  [0.0, 0.0188679245283, 0.0377358490566, 0.1132...             53   \n",
       "4  [0.0144927536232, 0.0144927536232, 0.028985507...             69   \n",
       "\n",
       "   Temperature ° F  Wind Speed mph  weather  \n",
       "0             73.9             3.5        3  \n",
       "1             73.0             3.5        3  \n",
       "2             72.0             3.5        3  \n",
       "3             72.0             3.5        3  \n",
       "4             71.1             4.6        3  "
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "start_out_df = start_out_df.merge(mete_df, how='left',on=['year','month','day','hour'])\n",
    "start_out_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "params = {\n",
    "    'rou': [0.7, 0.8],\n",
    "    'alpha': [0.6, 0.2, 0.05, 0.3, 0.1, 0.8],\n",
    "    'sigma': [start_out_df['Temperature ° F'].std(), start_out_df['Wind Speed mph'].std()],\n",
    "    'H': 48,\n",
    "    'J': 4,\n",
    "    'q': 1.2\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "def weather_sim(w1, w2):\n",
    "    ws = [\n",
    "        [1,  params['alpha'][0], params['alpha'][1], params['alpha'][2]],\n",
    "        [params['alpha'][0], 1, params['alpha'][3], params['alpha'][4]],\n",
    "        [params['alpha'][1], params['alpha'][3], 1, params['alpha'][5]],\n",
    "        [params['alpha'][2], params['alpha'][4], params['alpha'][5], 1],\n",
    "    ]\n",
    "    return ws[w1][w2]\n",
    "#     if w1 == w2:\n",
    "#         return 1\n",
    "#     if w1 > w2:\n",
    "#         return lam_2_func(w2, w1)\n",
    "#     # w1 < w2:\n",
    "#     s = [0, 3, 5]\n",
    "#     return params['alpha'][s[w1]+ w2 - w1 - 1]\n",
    "\n",
    "def sim_w(f1, f2):\n",
    "    if f1['isweekend'] != f2['isweekend']:\n",
    "        return 0\n",
    "    \n",
    "    delta_h = np.abs(f1['hour'] - f2['hour'])\n",
    "    if delta_h > 12:\n",
    "        delta_h = 24 - delta_h\n",
    "    delta_d = np.abs(f1['day'] - f2['day'])\n",
    "    lam_1 = (params['rou'][0] ** delta_h) * (params['rou'][1] ** delta_d)\n",
    "    \n",
    "    lam_2 = weather_sim(f1['weather'], f2['weather'])\n",
    "    \n",
    "    temp = (f1['Temperature ° F'] - f2['Temperature ° F']) / params['sigma'][0]\n",
    "    temp *= temp\n",
    "    wv = (f1['Wind Speed mph'] - f2['Wind Speed mph']) / params['sigma'][1]\n",
    "    wv *= wv\n",
    "    Ker = np.exp(-temp-wv)/(2*np.pi*params['sigma'][0]*params['sigma'][1])\n",
    "    \n",
    "    return lam_1*lam_2*Ker\n",
    "\n",
    "def predict_proportion(data, t):\n",
    "    w = [sim_w(data.loc[i], data.loc[t]) for i in range(t-params['H'], t)]\n",
    "    w = np.array([w])\n",
    "    w = w / w.sum()\n",
    "    Prop = [data.loc[i, 'proportion'] for i in range(t-params['H'], t)]\n",
    "    Prop = np.array(Prop)\n",
    "    return np.dot(w, Prop)\n",
    "\n",
    "def predict_proportion_ex(data, t):\n",
    "    e = [data.loc[i,'proportion'] - predict_proportion(data,i) for i in range(t-params['J'],t)]\n",
    "    e = np.array(e)\n",
    "    e = e.reshape(e.shape[0],e.shape[-1])\n",
    "    q = [params['q'] ** i for i in range(params['J'])]\n",
    "    q = np.array([q])\n",
    "    q = q / q.sum()\n",
    "    return predict_proportion(data, t)  + np.dot(q, e)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 0.02115446,  0.01786483,  0.09434252,  0.07854309,  0.07079415,\n",
       "         0.0437874 ,  0.04347242,  0.01668629,  0.00956986,  0.02953812,\n",
       "         0.0439952 ,  0.02745782,  0.0555485 ,  0.04223074,  0.04226319,\n",
       "         0.04828701,  0.03804634,  0.01216369,  0.09876508,  0.06164107,\n",
       "         0.02903719,  0.023776  ,  0.05103504]])"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predict_proportion_ex(start_out_df, 300)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 0.00497512,  0.00995025,  0.08789386,  0.07628524,  0.03565506,\n",
       "        0.04394693,  0.04892206,  0.01741294,  0.00497512,  0.03233831,\n",
       "        0.05638474,  0.0331675 ,  0.06135987,  0.04643449,  0.06799337,\n",
       "        0.04975124,  0.04394693,  0.00746269,  0.11442786,  0.06384743,\n",
       "        0.026534  ,  0.02072968,  0.04560531])"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "start_out_df.loc[300, 'proportion']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Inter-cluster Transition Learning\n",
    "\n",
    "$T_{t,C_i,C_j}$ is a transition probability from cluster $C_i$ to cluster $C_j$ in time $t$.\n",
    "\n",
    "$$T_{t,C_i,C_j} = Pr(\\text{check in at } C_j | \\text{ check out at } C_i \\text{ in time } t)$$\n",
    "\n",
    "\n",
    "Each row is a margin distribution.\n",
    "\n",
    "So the sum of each row in matrix T is 1. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>starttime</th>\n",
       "      <th>start station id</th>\n",
       "      <th>end station id</th>\n",
       "      <th>bcluster_x</th>\n",
       "      <th>bcluster_y</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>8/1/2014 0:00</td>\n",
       "      <td>470</td>\n",
       "      <td>312</td>\n",
       "      <td>18</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>8/1/2014 0:00</td>\n",
       "      <td>236</td>\n",
       "      <td>432</td>\n",
       "      <td>4</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>8/1/2014 0:00</td>\n",
       "      <td>224</td>\n",
       "      <td>340</td>\n",
       "      <td>9</td>\n",
       "      <td>22</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>8/1/2014 0:00</td>\n",
       "      <td>150</td>\n",
       "      <td>522</td>\n",
       "      <td>4</td>\n",
       "      <td>16</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>8/1/2014 0:00</td>\n",
       "      <td>519</td>\n",
       "      <td>477</td>\n",
       "      <td>13</td>\n",
       "      <td>14</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       starttime  start station id  end station id  bcluster_x  bcluster_y\n",
       "0  8/1/2014 0:00               470             312          18           3\n",
       "1  8/1/2014 0:00               236             432           4           4\n",
       "2  8/1/2014 0:00               224             340           9          22\n",
       "3  8/1/2014 0:00               150             522           4          16\n",
       "4  8/1/2014 0:00               519             477          13          14"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tmp_df = nyc_df[['starttime','start station id', 'end station id']]\n",
    "tmp_df = tmp_df.merge(station[['id', 'bcluster']],\n",
    "            how='left',left_on='start station id',right_on='id').drop('id',axis=1)\n",
    "tmp_df = tmp_df.merge(station[['id', 'bcluster']],\n",
    "            how='left',left_on='end station id',right_on='id').drop('id',axis=1)\n",
    "tmp_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>translation matrix</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">2014</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">8</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">1</th>\n",
       "      <th>0</th>\n",
       "      <td>[[0.428571428571, 0.142857142857, 0.0, 0.0, 0....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>[[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0,...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>[[0.833333333333, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>[[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>[[1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                           translation matrix\n",
       "2014 8 1 0  [[0.428571428571, 0.142857142857, 0.0, 0.0, 0....\n",
       "         1  [[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0,...\n",
       "         2  [[0.833333333333, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0...\n",
       "         3  [[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,...\n",
       "         4  [[1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,..."
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tmp_df['sdt'] = tmp_df['starttime'].map(lambda x:datetime.strptime(x, '%m/%d/%Y %H:%M'))\n",
    "tmp_df['year'] = tmp_df['sdt'].map(lambda x:x.year)\n",
    "tmp_df['month'] = tmp_df['sdt'].map(lambda x:x.month)\n",
    "tmp_df['day'] = tmp_df['sdt'].map(lambda x:x.day)\n",
    "tmp_df['hour'] = tmp_df['sdt'].map(lambda x:x.hour)\n",
    "\n",
    "by_ymdh = dict(tmp_df.groupby(['year','month','day','hour']).groups)\n",
    "tranmat = {}\n",
    "for key in by_ymdh.keys():\n",
    "    start = dict(tmp_df.loc[by_ymdh[key]].groupby('bcluster_x').groups)\n",
    "    for s_key in start.keys():\n",
    "        end = dict(tmp_df.loc[start[s_key]].groupby('bcluster_y').size())\n",
    "        count = np.array(list(map(lambda x:end[x] if x in end else 0, range(K1))))\n",
    "        start[s_key] = count / count.sum()\n",
    "    if len(start) < K1:\n",
    "        for i in range(K1):\n",
    "            if i not in start.keys():\n",
    "                start[i] = np.zeros(K1)\n",
    "    tranmat[key] = np.array(list(start.values()))\n",
    "tranmat_df = pd.DataFrame({'translation matrix': pd.Series(tranmat)})\n",
    "tranmat_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>level_0</th>\n",
       "      <th>level_1</th>\n",
       "      <th>level_2</th>\n",
       "      <th>level_3</th>\n",
       "      <th>translation matrix</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>[[0.428571428571, 0.142857142857, 0.0, 0.0, 0....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>[[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0,...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>[[0.833333333333, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>[[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>[[1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   level_0  level_1  level_2  level_3  \\\n",
       "0     2014        8        1        0   \n",
       "1     2014        8        1        1   \n",
       "2     2014        8        1        2   \n",
       "3     2014        8        1        3   \n",
       "4     2014        8        1        4   \n",
       "\n",
       "                                  translation matrix  \n",
       "0  [[0.428571428571, 0.142857142857, 0.0, 0.0, 0....  \n",
       "1  [[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0,...  \n",
       "2  [[0.833333333333, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0...  \n",
       "3  [[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,...  \n",
       "4  [[1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,...  "
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tranmat_df = tranmat_df.reset_index()\n",
    "tranmat_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>year</th>\n",
       "      <th>month</th>\n",
       "      <th>day</th>\n",
       "      <th>hour</th>\n",
       "      <th>translation matrix</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>[[0.428571428571, 0.142857142857, 0.0, 0.0, 0....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>[[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0,...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>[[0.833333333333, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>[[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>[[1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   year  month  day  hour                                 translation matrix\n",
       "0  2014      8    1     0  [[0.428571428571, 0.142857142857, 0.0, 0.0, 0....\n",
       "1  2014      8    1     1  [[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0,...\n",
       "2  2014      8    1     2  [[0.833333333333, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0...\n",
       "3  2014      8    1     3  [[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,...\n",
       "4  2014      8    1     4  [[1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,..."
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tranmat_df = tranmat_df.rename(columns={'level_0':'year', 'level_1':'month', 'level_2':'day', 'level_3':'hour'})\n",
    "tranmat_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 0.42857143,  0.14285714,  0.        ,  0.        ,  0.14285714,\n",
       "         0.        ,  0.        ,  0.        ,  0.14285714,  0.        ,\n",
       "         0.        ,  0.        ,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.14285714,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.        ],\n",
       "       [ 0.        ,  0.        ,  0.        ,  0.04545455,  0.36363636,\n",
       "         0.13636364,  0.09090909,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.        ,  0.09090909,  0.        ,\n",
       "         0.04545455,  0.        ,  0.        ,  0.        ,  0.09090909,\n",
       "         0.04545455,  0.        ,  0.09090909],\n",
       "       [ 0.03030303,  0.        ,  0.09090909,  0.09090909,  0.18181818,\n",
       "         0.09090909,  0.12121212,  0.03030303,  0.        ,  0.03030303,\n",
       "         0.        ,  0.03030303,  0.        ,  0.        ,  0.03030303,\n",
       "         0.        ,  0.03030303,  0.03030303,  0.09090909,  0.06060606,\n",
       "         0.        ,  0.        ,  0.06060606],\n",
       "       [ 0.        ,  0.        ,  0.11428571,  0.02857143,  0.25714286,\n",
       "         0.05714286,  0.08571429,  0.        ,  0.        ,  0.02857143,\n",
       "         0.02857143,  0.        ,  0.        ,  0.02857143,  0.08571429,\n",
       "         0.02857143,  0.08571429,  0.        ,  0.02857143,  0.05714286,\n",
       "         0.        ,  0.02857143,  0.05714286],\n",
       "       [ 0.        ,  0.        ,  0.11764706,  0.11764706,  0.11764706,\n",
       "         0.        ,  0.        ,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.        ,  0.05882353,  0.05882353,\n",
       "         0.23529412,  0.        ,  0.        ,  0.11764706,  0.11764706,\n",
       "         0.05882353,  0.        ,  0.        ],\n",
       "       [ 0.        ,  0.        ,  0.04166667,  0.16666667,  0.25      ,\n",
       "         0.        ,  0.04166667,  0.        ,  0.        ,  0.        ,\n",
       "         0.04166667,  0.04166667,  0.        ,  0.        ,  0.04166667,\n",
       "         0.04166667,  0.        ,  0.        ,  0.20833333,  0.08333333,\n",
       "         0.        ,  0.        ,  0.04166667],\n",
       "       [ 0.28571429,  0.42857143,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.14285714,  0.14285714,  0.        ,\n",
       "         0.        ,  0.        ,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.        ],\n",
       "       [ 0.        ,  0.33333333,  0.        ,  0.        ,  0.66666667,\n",
       "         0.        ,  0.        ,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.        ],\n",
       "       [ 0.08333333,  0.        ,  0.        ,  0.        ,  0.        ,\n",
       "         0.16666667,  0.        ,  0.16666667,  0.        ,  0.08333333,\n",
       "         0.16666667,  0.16666667,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.        ,  0.08333333,  0.        ,\n",
       "         0.        ,  0.        ,  0.08333333],\n",
       "       [ 0.        ,  0.        ,  0.07692308,  0.07692308,  0.        ,\n",
       "         0.        ,  0.30769231,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.07692308,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.        ,  0.15384615,  0.07692308,\n",
       "         0.07692308,  0.        ,  0.15384615],\n",
       "       [ 0.        ,  0.        ,  0.        ,  0.16666667,  0.16666667,\n",
       "         0.        ,  0.        ,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.66666667,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.        ],\n",
       "       [ 0.        ,  0.        ,  0.05555556,  0.05555556,  0.        ,\n",
       "         0.05555556,  0.        ,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.11111111,  0.05555556,  0.05555556,\n",
       "         0.        ,  0.16666667,  0.        ,  0.22222222,  0.        ,\n",
       "         0.22222222,  0.        ,  0.        ],\n",
       "       [ 0.        ,  0.        ,  0.0625    ,  0.        ,  0.0625    ,\n",
       "         0.25      ,  0.        ,  0.        ,  0.0625    ,  0.        ,\n",
       "         0.        ,  0.        ,  0.        ,  0.125     ,  0.125     ,\n",
       "         0.        ,  0.1875    ,  0.        ,  0.0625    ,  0.        ,\n",
       "         0.0625    ,  0.        ,  0.        ],\n",
       "       [ 0.        ,  0.        ,  0.11111111,  0.        ,  0.05555556,\n",
       "         0.11111111,  0.        ,  0.        ,  0.        ,  0.05555556,\n",
       "         0.05555556,  0.        ,  0.11111111,  0.05555556,  0.11111111,\n",
       "         0.        ,  0.05555556,  0.        ,  0.11111111,  0.        ,\n",
       "         0.11111111,  0.        ,  0.05555556],\n",
       "       [ 0.        ,  0.        ,  0.17647059,  0.05882353,  0.11764706,\n",
       "         0.05882353,  0.05882353,  0.        ,  0.05882353,  0.        ,\n",
       "         0.        ,  0.        ,  0.11764706,  0.05882353,  0.        ,\n",
       "         0.05882353,  0.        ,  0.        ,  0.17647059,  0.        ,\n",
       "         0.        ,  0.        ,  0.05882353],\n",
       "       [ 0.        ,  0.        ,  0.        ,  0.        ,  0.        ,\n",
       "         0.14285714,  0.        ,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.14285714,  0.28571429,  0.        ,\n",
       "         0.28571429,  0.        ,  0.        ,  0.        ,  0.        ,\n",
       "         0.14285714,  0.        ,  0.        ],\n",
       "       [ 0.22222222,  0.11111111,  0.        ,  0.        ,  0.11111111,\n",
       "         0.        ,  0.11111111,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.44444444,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.        ],\n",
       "       [ 0.        ,  0.        ,  0.06666667,  0.02222222,  0.04444444,\n",
       "         0.        ,  0.06666667,  0.02222222,  0.        ,  0.02222222,\n",
       "         0.        ,  0.02222222,  0.06666667,  0.06666667,  0.06666667,\n",
       "         0.02222222,  0.02222222,  0.        ,  0.33333333,  0.11111111,\n",
       "         0.04444444,  0.        ,  0.        ],\n",
       "       [ 0.        ,  0.        ,  0.16393443,  0.09836066,  0.14754098,\n",
       "         0.04918033,  0.09836066,  0.        ,  0.        ,  0.01639344,\n",
       "         0.01639344,  0.        ,  0.04918033,  0.        ,  0.01639344,\n",
       "         0.01639344,  0.        ,  0.        ,  0.14754098,  0.1147541 ,\n",
       "         0.06557377,  0.        ,  0.        ],\n",
       "       [ 0.        ,  0.        ,  0.        ,  0.        ,  0.09090909,\n",
       "         0.        ,  0.        ,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.09090909,  0.        ,  0.09090909,\n",
       "         0.        ,  0.        ,  0.        ,  0.36363636,  0.        ,\n",
       "         0.27272727,  0.        ,  0.09090909],\n",
       "       [ 0.        ,  0.        ,  0.        ,  0.        ,  0.5       ,\n",
       "         0.        ,  0.        ,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.5       ,  0.        ],\n",
       "       [ 0.1       ,  0.        ,  0.        ,  0.1       ,  0.        ,\n",
       "         0.        ,  0.1       ,  0.        ,  0.        ,  0.1       ,\n",
       "         0.2       ,  0.        ,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.1       ,  0.2       ,  0.        ,\n",
       "         0.        ,  0.        ,  0.1       ],\n",
       "       [ 0.        ,  0.        ,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.        ,  0.        ,  0.        ,\n",
       "         0.        ,  0.        ,  0.        ]])"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tranmat_df.loc[0, 'translation matrix']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Trip Duration Learning"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![](figure/lognormdist.png)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>bcluster_x</th>\n",
       "      <th>bcluster_y</th>\n",
       "      <th>tripduration</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>18</td>\n",
       "      <td>3</td>\n",
       "      <td>1142</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>4</td>\n",
       "      <td>4</td>\n",
       "      <td>117</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>9</td>\n",
       "      <td>22</td>\n",
       "      <td>546</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>16</td>\n",
       "      <td>2126</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>13</td>\n",
       "      <td>14</td>\n",
       "      <td>329</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   bcluster_x  bcluster_y  tripduration\n",
       "0          18           3          1142\n",
       "1           4           4           117\n",
       "2           9          22           546\n",
       "3           4          16          2126\n",
       "4          13          14           329"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tmp_df = nyc_df[['tripduration','start station id', 'end station id']]\n",
    "tmp_df = tmp_df.merge(station[['id', 'bcluster']],\n",
    "            how='left',left_on='start station id',right_on='id').drop('id',axis=1)\n",
    "tmp_df = tmp_df.merge(station[['id', 'bcluster']],\n",
    "            how='left',left_on='end station id',right_on='id').drop('id',axis=1)\n",
    "\n",
    "tmp_df = tmp_df[['bcluster_x', 'bcluster_y', 'tripduration']]\n",
    "tmp_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>duration</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">0</th>\n",
       "      <th>0</th>\n",
       "      <td>[2099, 2068, 1620, 1737, 1345, 813, 248, 197, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>[271, 246, 658, 290, 391, 570, 230, 484, 492, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>[1887, 2075, 1546, 1667, 1668, 2070, 1982, 194...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>[1604, 1079, 2229, 1704, 1860, 1817, 1646, 182...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>[1556, 2203, 1529, 1727, 1604, 2021, 2053, 151...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                              duration\n",
       "0 0  [2099, 2068, 1620, 1737, 1345, 813, 248, 197, ...\n",
       "  1  [271, 246, 658, 290, 391, 570, 230, 484, 492, ...\n",
       "  2  [1887, 2075, 1546, 1667, 1668, 2070, 1982, 194...\n",
       "  3  [1604, 1079, 2229, 1704, 1860, 1817, 1646, 182...\n",
       "  4  [1556, 2203, 1529, 1727, 1604, 2021, 2053, 151..."
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dura_by_cluster = dict(tmp_df.groupby(['bcluster_x', 'bcluster_y']).groups)\n",
    "for key in dura_by_cluster:\n",
    "    dura_by_cluster[key] = np.array(list(tmp_df.loc[dura_by_cluster[key], 'tripduration']))\n",
    "\n",
    "dura_df = pd.DataFrame({'duration': pd.Series(dura_by_cluster)})\n",
    "dura_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>duration</th>\n",
       "      <th>logx</th>\n",
       "      <th>mu</th>\n",
       "      <th>sigma</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">0</th>\n",
       "      <th>0</th>\n",
       "      <td>[2099, 2068, 1620, 1737, 1345, 813, 248, 197, ...</td>\n",
       "      <td>[7.64921631982, 7.63433723563, 7.39018142823, ...</td>\n",
       "      <td>6.011114</td>\n",
       "      <td>0.769506</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>[271, 246, 658, 290, 391, 570, 230, 484, 492, ...</td>\n",
       "      <td>[5.60211882088, 5.50533153593, 6.48920493133, ...</td>\n",
       "      <td>6.075236</td>\n",
       "      <td>0.511757</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>[1887, 2075, 1546, 1667, 1668, 2070, 1982, 194...</td>\n",
       "      <td>[7.54274354537, 7.63771643266, 7.34342622915, ...</td>\n",
       "      <td>7.563000</td>\n",
       "      <td>0.165129</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>[1604, 1079, 2229, 1704, 1860, 1817, 1646, 182...</td>\n",
       "      <td>[7.38025578843, 6.98378996526, 7.70930833339, ...</td>\n",
       "      <td>7.383190</td>\n",
       "      <td>0.185231</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>[1556, 2203, 1529, 1727, 1604, 2021, 2053, 151...</td>\n",
       "      <td>[7.34987370474, 7.6975753468, 7.33236920593, 7...</td>\n",
       "      <td>7.422137</td>\n",
       "      <td>0.150796</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                              duration  \\\n",
       "0 0  [2099, 2068, 1620, 1737, 1345, 813, 248, 197, ...   \n",
       "  1  [271, 246, 658, 290, 391, 570, 230, 484, 492, ...   \n",
       "  2  [1887, 2075, 1546, 1667, 1668, 2070, 1982, 194...   \n",
       "  3  [1604, 1079, 2229, 1704, 1860, 1817, 1646, 182...   \n",
       "  4  [1556, 2203, 1529, 1727, 1604, 2021, 2053, 151...   \n",
       "\n",
       "                                                  logx        mu     sigma  \n",
       "0 0  [7.64921631982, 7.63433723563, 7.39018142823, ...  6.011114  0.769506  \n",
       "  1  [5.60211882088, 5.50533153593, 6.48920493133, ...  6.075236  0.511757  \n",
       "  2  [7.54274354537, 7.63771643266, 7.34342622915, ...  7.563000  0.165129  \n",
       "  3  [7.38025578843, 6.98378996526, 7.70930833339, ...  7.383190  0.185231  \n",
       "  4  [7.34987370474, 7.6975753468, 7.33236920593, 7...  7.422137  0.150796  "
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dura_df['logx'] = dura_df['duration'].map(lambda x:np.log(x))\n",
    "dura_df['mu'] = dura_df['logx'].map(lambda x:x.mean())\n",
    "dura_df['sigma'] = dura_df['logx'].map(lambda x:x.std())\n",
    "dura_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[]"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "missing = [(x, y)\n",
    "           for x in range(K1) for y in range(K1)\n",
    "           if (x,y) not in dura_by_cluster.keys()]\n",
    "missing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[ 6.01111407  6.07523604  7.56299966  7.38319     7.42213696  7.78540476\n",
      "   7.54336015  6.90178716  6.5142956   7.35346566  7.44502963  7.38645155\n",
      "   8.46596128  7.67972835  7.78450878  7.77686943  7.76072224  6.48106355\n",
      "   7.82434405  7.76229406  8.32360844  7.46803452  7.34003182]\n",
      " [ 6.17700849  5.86937646  7.43915199  7.21745371  7.34919298  7.47406198\n",
      "   7.44199043  6.50553714  6.05933739  7.18087515  7.11218205  7.28399285\n",
      "   7.77450548  7.74967203  7.70402252  7.53209015  7.67082483  6.77821097\n",
      "   7.67326053  7.49929648  7.80998294  7.4428878   7.06322448]\n",
      " [ 7.59956714  7.54957409  5.86262371  6.32886902  6.21852126  6.05799657\n",
      "   6.03106068  7.46515962  7.4921674   6.96071605  6.92211905  7.25566763\n",
      "   7.22679296  6.70821593  6.55525658  6.23650813  7.09077095  7.41196663\n",
      "   6.2388953   6.37227333  7.03890839  6.8415654   6.77304124]\n",
      " [ 7.43997418  7.2744042   6.24069232  5.93943554  6.17611065  6.56340529\n",
      "   6.10532858  7.28770067  7.2163776   6.49483842  6.42236362  6.92279217\n",
      "   7.52304535  7.12457577  7.1092279   6.82812304  7.31851959  7.14747282\n",
      "   6.84657403  6.54827049  7.38674813  6.42715146  6.1977901 ]\n",
      " [ 7.51999881  7.43072163  6.20198308  6.1809585   5.89056063  6.17500524\n",
      "   6.44746423  7.35146573  7.44139537  7.02708084  6.95892127  7.1759579\n",
      "   7.44524632  6.98130727  7.11014287  6.59876536  7.09836889  7.13815177\n",
      "   6.85430114  6.76239275  7.42416001  6.28997592  6.44371594]\n",
      " [ 7.62061356  7.54911728  6.05558838  6.57508014  6.29873323  5.85842877\n",
      "   6.59133725  7.43777169  7.51967415  7.21815089  7.25174328  7.35810533\n",
      "   7.13201125  6.49059352  6.52340106  5.93081423  6.75790029  7.39805361\n",
      "   6.497314    6.84009563  7.0413775   6.96770288  6.91023031]\n",
      " [ 7.63933234  7.41655764  6.01583598  6.14230645  6.46857771  6.6545666\n",
      "   5.83872133  7.49144451  7.43373659  6.89493398  6.61577845  7.14770783\n",
      "   7.30509438  6.94591453  6.79499325  6.70783587  7.36340729  7.45851003\n",
      "   6.30773523  6.02108607  7.21044853  6.91422419  6.75915432]\n",
      " [ 6.88435664  6.44299287  7.4482466   7.20679608  7.31115973  7.58071056\n",
      "   7.59711169  6.38057282  6.28410445  7.20503024  7.02337971  7.21336557\n",
      "   8.01436053  7.72210873  7.88329199  7.60036261  7.75996351  6.95124468\n",
      "   7.61835759  7.48661792  8.03395616  7.13042308  7.01654573]\n",
      " [ 6.68597726  6.07206818  7.36439023  7.20367047  7.32012321  7.60771923\n",
      "   7.33032699  6.42676319  6.37732387  7.25402638  7.18531959  7.05616998\n",
      "   7.8182059   7.6113999   7.82490021  7.4114926   7.7317494   6.92828135\n",
      "   7.61666158  7.6698936   7.88144802  7.40302939  7.06309628]\n",
      " [ 7.40787415  7.2497232   7.00617388  6.50735178  6.95225498  7.26965218\n",
      "   6.90593625  7.33542545  7.15218671  6.12503008  6.09342388  6.16946999\n",
      "   7.78843633  7.43636932  7.55852369  7.30781029  7.6358881   7.44948411\n",
      "   7.27933197  7.01419146  7.49585884  6.77504536  6.25940684]\n",
      " [ 7.45926088  7.20645021  6.95414129  6.46370356  7.00544697  7.27697345\n",
      "   6.56892087  7.19163829  7.29043523  6.08730609  5.87805911  6.40384827\n",
      "   7.5591839   7.44263101  7.33108697  7.32689383  7.67879572  7.31433622\n",
      "   7.02954957  6.60249099  7.3199109   6.99706629  6.57024646]\n",
      " [ 7.57753042  7.33273521  7.35878174  7.00954988  7.20458015  7.38980393\n",
      "   7.14266274  7.42289319  7.28218154  6.32046504  6.36915708  6.13567002\n",
      "   7.79165997  7.58533512  7.6063303   7.48536546  7.64801631  7.47377001\n",
      "   7.3545665   7.13522533  7.51854087  7.04838048  6.71928811]\n",
      " [ 7.93293704  7.83013805  7.13874813  7.46670588  7.44003134  7.1384078\n",
      "   7.28316169  7.98600421  7.91277803  7.64927996  7.48753898  7.71148366\n",
      "   6.40733355  6.41331678  6.45514858  6.96173021  6.48924332  8.01781355\n",
      "   6.91431314  7.22573683  6.31033625  7.55303652  7.62160181]\n",
      " [ 7.75945078  7.71301207  6.59578767  7.05247702  6.94695251  6.26429865\n",
      "   6.90394918  7.65658399  7.68994114  7.47791677  7.40296507  7.59346011\n",
      "   6.48183396  5.88468735  6.23908379  6.16494016  6.2685793   7.65934016\n",
      "   6.69367774  7.10262232  6.51042738  7.29330352  7.30019061]\n",
      " [ 7.79256366  7.79450141  6.46553913  7.14475026  7.08218665  6.51037647\n",
      "   6.84553533  7.87604561  7.82215815  7.47995121  7.32784237  7.57191166\n",
      "   6.41655375  6.21302789  5.90647475  6.23012022  6.72527134  7.69269122\n",
      "   6.32102681  6.83001899  6.1601633   7.3658894   7.34927418]\n",
      " [ 7.69975392  7.61984069  6.22672939  6.88078989  6.69291032  5.87049993\n",
      "   6.68427907  7.69194415  7.6257063   7.37722182  7.28195453  7.47292091\n",
      "   7.05137313  6.27172712  6.26631086  5.98772718  6.6723584   7.64814497\n",
      "   6.39225408  6.86575915  6.87202475  7.16287148  7.14623173]\n",
      " [ 7.81273923  7.77273687  7.02051251  7.27195141  7.14868225  6.65530039\n",
      "   7.26923318  7.89024465  7.70962975  7.61698644  7.63258791  7.74763663\n",
      "   6.66509124  6.24801579  6.77299757  6.65827079  6.26500353  7.72742965\n",
      "   7.13510367  7.51220805  6.93610801  7.38637914  7.37482877]\n",
      " [ 6.59116231  6.79130535  7.38787675  7.10497543  7.12956553  7.44122257\n",
      "   7.22833308  7.06702931  6.91485551  7.36727252  7.31134319  7.49114506\n",
      "   8.00691272  7.71275825  7.63761406  7.51873916  7.75753813  6.42975488\n",
      "   7.60248257  7.37941755  7.72613053  7.10868281  7.04788434]\n",
      " [ 7.75996487  7.70988537  6.25370021  6.85971646  6.89658719  6.46584067\n",
      "   6.38688042  7.60360462  7.60604809  7.27952586  7.06559619  7.33990232\n",
      "   6.84186859  6.70269958  6.27559714  6.37907976  7.0692593   7.60886334\n",
      "   5.82698403  6.19462937  6.64078635  7.31907556  7.23227325]\n",
      " [ 7.68561075  7.54911986  6.43878973  6.5639041   6.8016297   6.85372099\n",
      "   6.0647351   7.60969186  7.60542816  7.08049701  6.63063062  7.12936593\n",
      "   7.26552716  7.13238759  6.84634021  6.93094785  7.48802215  7.54052584\n",
      "   6.17630901  5.98099669  7.00532262  7.14516301  6.9955753 ]\n",
      " [ 7.86949976  7.92507079  6.98082437  7.33899652  7.36310515  6.99927498\n",
      "   7.14849763  7.7919735   8.13369368  7.50987128  7.30152976  7.47120446\n",
      "   6.44209049  6.57798681  6.18194883  6.88653824  6.91561665  7.75974067\n",
      "   6.49262859  6.92313083  6.32103226  7.77815679  7.52352372]\n",
      " [ 7.65806327  7.31746413  6.81948655  6.39184037  6.29939516  6.85584235\n",
      "   6.8482097   7.21301886  7.18002699  6.73417055  6.77771261  6.97574249\n",
      "   7.63428127  7.2804302   7.34959856  7.08626186  7.24484094  7.02591316\n",
      "   7.23541359  7.07331212  7.61358158  6.27862653  6.08763661]\n",
      " [ 7.32263274  7.09175033  6.74942305  6.17289958  6.38670951  6.85396402\n",
      "   6.69744812  7.07786286  7.00796163  6.28694412  6.52031657  6.65529115\n",
      "   7.70074057  7.30822821  7.36687321  7.06731068  7.29839315  7.03109614\n",
      "   7.17862799  6.95415755  7.59566125  6.10025597  5.95786771]]\n",
      "[[ 0.76950639  0.51175733  0.16512937  0.18523112  0.15079584  0.36659937\n",
      "   0.21314235  0.43229166  0.45679173  0.34455871  0.19631405  0.18374919\n",
      "   0.62840692  0.154388    0.13856909  0.24660724  0.16791422  0.73201785\n",
      "   0.15336465  0.1503563   0.          0.28557429  0.28620882]\n",
      " [ 0.49867314  0.67328426  0.22979897  0.29858712  0.37162347  0.25071003\n",
      "   0.19776601  0.56526471  0.61930344  0.34394999  0.28276851  0.33320458\n",
      "   0.07650329  0.3033406   0.13102827  0.17589885  0.13936309  0.59936039\n",
      "   0.22683038  0.15270897  0.1639696   0.24371903  0.27121281]\n",
      " [ 0.18170584  0.25157619  0.78127717  0.536182    0.50135368  0.45304298\n",
      "   0.53453841  0.32780343  0.40100712  0.39161092  0.39757731  0.37662467\n",
      "   0.39086784  0.40589037  0.50937087  0.45837206  0.3795656   0.25428565\n",
      "   0.4783327   0.43190621  0.38069695  0.39209314  0.39906429]\n",
      " [ 0.19260978  0.25715756  0.56864172  0.78218129  0.53610674  0.4262814\n",
      "   0.55570094  0.37829663  0.30932182  0.50619992  0.56071511  0.39536329\n",
      "   0.3391465   0.39384014  0.32954406  0.38128592  0.33950055  0.3280124\n",
      "   0.37637954  0.4659638   0.32489196  0.48639875  0.57084665]\n",
      " [ 0.21567021  0.25821079  0.49392246  0.52476369  0.8103888   0.52073223\n",
      "   0.34423646  0.38799988  0.3476731   0.43016442  0.38061071  0.29888941\n",
      "   0.33620626  0.33564976  0.30786052  0.41666999  0.3230249   0.34151693\n",
      "   0.33834611  0.32771873  0.30691628  0.73653695  0.49358676]\n",
      " [ 0.13000712  0.23634373  0.4542752   0.39505762  0.49927806  0.80023391\n",
      "   0.31966325  0.27318708  0.29671091  0.28750004  0.34116466  0.27117158\n",
      "   0.29813077  0.41979601  0.41701345  0.61871683  0.42943022  0.28064971\n",
      "   0.38702596  0.35335406  0.40515309  0.40871432  0.40061105]\n",
      " [ 0.15927878  0.19060595  0.52010902  0.52782498  0.40118807  0.31174697\n",
      "   0.84140221  0.43681498  0.30481368  0.37819154  0.47898383  0.34106405\n",
      "   0.39142812  0.28662541  0.37961421  0.36756134  0.36064383  0.35907403\n",
      "   0.50584525  0.61244948  0.31733795  0.3819536   0.39182873]\n",
      " [ 0.39870644  0.59822134  0.37226651  0.33211164  0.36673132  0.35825378\n",
      "   0.67987297  0.93628484  0.66562742  0.40261111  0.41444859  0.48801781\n",
      "   0.45742334  0.37340591  0.49906151  0.21784004  0.31713021  0.67613055\n",
      "   0.37207237  0.30410443  0.5450437   0.29920606  0.3886845 ]\n",
      " [ 0.55108087  0.66822721  0.21627954  0.30772089  0.25036885  0.72524062\n",
      "   0.23572962  0.70929327  1.03505194  0.38630039  0.39615048  0.26789256\n",
      "   0.1377075   0.19506751  0.3875195   0.32183263  0.27631806  0.63904564\n",
      "   0.18595853  0.19343717  0.19800635  0.34074074  0.29551269]\n",
      " [ 0.18340618  0.45487813  0.36120604  0.52209545  0.36191135  0.28881166\n",
      "   0.40671646  0.43342818  0.35921092  1.00743328  0.63349727  0.68650355\n",
      "   0.49242173  0.33882032  0.52518701  0.29077492  0.42349731  0.43831364\n",
      "   0.30013629  0.39981794  0.23843927  0.41907345  0.56255818]\n",
      " [ 0.18849622  0.26172057  0.39701605  0.54423767  0.36262231  0.33539859\n",
      "   0.48921469  0.42896072  0.43414357  0.58724383  0.84296135  0.5819332\n",
      "   0.34635776  0.30922404  0.33860987  0.27664291  0.35590122  0.26434514\n",
      "   0.31786416  0.48288162  0.27620776  0.40370965  0.44699594]\n",
      " [ 0.17898621  0.32438719  0.31323567  0.42625612  0.30844067  0.27330974\n",
      "   0.35675845  0.44611823  0.39681377  0.79624444  0.63046373  0.96117359\n",
      "   0.44127517  0.32125414  0.33224836  0.26288788  0.26381779  0.31439385\n",
      "   0.28336916  0.40439095  0.25056608  0.38242283  0.45044841]\n",
      " [ 0.16567717  0.22728297  0.33624593  0.28809252  0.30142703  0.31263843\n",
      "   0.30699235  0.35865564  0.13242328  0.30734236  0.26290569  0.33078061\n",
      "   1.12488241  0.50322747  0.51420531  0.39885258  0.68968803  0.40489919\n",
      "   0.37418924  0.34091463  0.64305708  0.17644859  0.24064833]\n",
      " [ 0.14568794  0.17789354  0.42225688  0.37145824  0.33676599  0.47824212\n",
      "   0.3060689   0.22464116  0.14463938  0.37032723  0.28316874  0.36536089\n",
      "   0.53226137  0.78221843  0.3952839   0.51764509  0.60151078  0.27773613\n",
      "   0.36438032  0.34453847  0.4211553   0.26726108  0.3286773 ]\n",
      " [ 0.15737617  0.25315547  0.51195247  0.32521828  0.28928593  0.39751168\n",
      "   0.39194067  0.38257319  0.35903276  0.37332026  0.2888367   0.30921958\n",
      "   0.56493947  0.424666    0.81299571  0.43537143  0.3863511   0.28749631\n",
      "   0.49812742  0.39126813  0.55630819  0.28218877  0.2833215 ]\n",
      " [ 0.11553982  0.24162058  0.52753132  0.41443361  0.36372452  0.61250577\n",
      "   0.40435521  0.22278022  0.20475149  0.31170751  0.36387151  0.31246113\n",
      "   0.4105424   0.47441658  0.53086135  0.89980395  0.49232663  0.29780424\n",
      "   0.47545765  0.38465191  0.4414378   0.30567287  0.37767627]\n",
      " [ 0.15175676  0.31716996  0.32673171  0.32974855  0.37088784  0.42789019\n",
      "   0.29258668  0.40051411  0.11951901  0.29247157  0.28410665  0.37685146\n",
      "   0.70985592  0.56193148  0.36216126  0.44610971  1.03656223  0.27598775\n",
      "   0.31553785  0.4085022   0.41078588  0.26548988  0.31291044]\n",
      " [ 0.70079507  0.52856382  0.2941583   0.30810723  0.37497733  0.24608003\n",
      "   0.32855887  0.61718029  0.53015202  0.27328323  0.36403826  0.40635272\n",
      "   0.53152917  0.47370041  0.1880226   0.22174383  0.3597484   0.97805731\n",
      "   0.24612482  0.33329916  0.15482606  0.40179483  0.37691032]\n",
      " [ 0.15384746  0.16318922  0.48536865  0.34657442  0.34996205  0.38180455\n",
      "   0.49445737  0.23253228  0.26116863  0.32481862  0.34586375  0.2843518\n",
      "   0.43029922  0.33972421  0.48210689  0.42365942  0.35568333  0.32565173\n",
      "   0.6189508   0.55811071  0.51219281  0.32855584  0.35978716]\n",
      " [ 0.13866566  0.26685044  0.41166856  0.45489765  0.33620509  0.37130553\n",
      "   0.61275702  0.4561826   0.22443131  0.38143558  0.49883998  0.36622928\n",
      "   0.41560864  0.36345626  0.37052917  0.3438008   0.37299812  0.38213049\n",
      "   0.53395552  0.88885273  0.40082623  0.32944919  0.35810115]\n",
      " [ 0.21426801  0.195715    0.37387032  0.29581     0.28136227  0.41554336\n",
      "   0.29303211  0.21607067  0.34795121  0.35471272  0.26415936  0.26971824\n",
      "   0.66255311  0.42005852  0.57514216  0.37911667  0.39677421  0.25984189\n",
      "   0.48445204  0.40157051  0.98692718  0.569466    0.35749739]\n",
      " [ 0.17789208  0.27509284  0.36886543  0.45556063  0.64132734  0.37498588\n",
      "   0.36660012  0.319453    0.24368134  0.43691052  0.32642868  0.34324182\n",
      "   0.28000599  0.31621815  0.22400678  0.32119416  0.25173211  0.35621807\n",
      "   0.2746953   0.27295946  0.10189961  1.07168035  0.56052336]\n",
      " [ 0.23864781  0.29728905  0.38661952  0.58143432  0.53794304  0.36906516\n",
      "   0.41001945  0.4332283   0.35401687  0.58936841  0.45227143  0.46196632\n",
      "   0.37922543  0.30732492  0.39957694  0.29629913  0.24473543  0.37144753\n",
      "   0.26838283  0.3542151   0.27309388  0.58902044  0.82433427]]\n"
     ]
    }
   ],
   "source": [
    "dura_df['size'] = dura_df['duration'].map(lambda x:len(x))\n",
    "mean_of_mu, mean_of_sigma = dura_df[dura_df['size'] > 5][['mu','sigma']].mean()\n",
    "Mu, Sigma = np.zeros((K1,K1)), np.zeros((K1,K1))\n",
    "for i in range(K1):\n",
    "    for j in range(K1):\n",
    "        try:\n",
    "            Mu[i, j] = dura_df.loc[(i,j), 'mu']\n",
    "            Sigma[i, j] = dura_df.loc[(i,j), 'sigma']\n",
    "        except:\n",
    "            Mu[i, j], Sigma[i, j] = mean_of_mu, mean_of_sigma\n",
    "print(Mu)\n",
    "print(Sigma)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "duration    [2246, 3536, 2113, 2106, 1755, 2029, 2104, 183...\n",
       "logx        [7.7169061353, 8.17075142376, 7.65586401762, 7...\n",
       "mu                                                    7.69269\n",
       "sigma                                                0.287496\n",
       "size                                                       68\n",
       "Name: (14, 17), dtype: object"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dura_df.loc[(14,17)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.28749631069489373"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# if there is only one sample, then sigma will be zero\n",
    "# this is caused by unstable biclustering\n",
    "for i in range(K1):\n",
    "    for j in range(K1):\n",
    "        if np.abs(Sigma[i, j]) < 1e-16:\n",
    "            Sigma[i, j] = mean_of_sigma\n",
    "Sigma[14, 17]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Online process\n",
    "## Cluster Check-out Inference"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "995.156 1206\n"
     ]
    }
   ],
   "source": [
    "t = 300\n",
    "train = start_out_df.drop('proportion',axis=1).head(t - 1)\n",
    "train_X = train.drop('entiretraffic',axis=1)\n",
    "train_y = train['entiretraffic']\n",
    "test = start_out_df.drop('proportion',axis=1).loc[[t]]\n",
    "test_X = test.drop('entiretraffic',axis=1)\n",
    "test_y = test['entiretraffic']\n",
    "dtrain = xgb.DMatrix(train_X, train_y)\n",
    "dtest = xgb.DMatrix(test_X)\n",
    "model = xgb.train(xgb_params, dtrain, num_boost_round=1000)\n",
    "Et = model.predict(dtest)[0]\n",
    "print(Et, start_out_df.loc[t,'entiretraffic'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[   5.13391285    9.9980307   120.33162018   81.83813299   38.87987472\n",
      "    38.90532983   50.82672766   20.20538575    4.68842551   39.53372182\n",
      "    64.72610396   45.29122193   78.64970228   69.69158973   51.91094117\n",
      "    50.72833683   56.86544025    5.83055549  125.67629699   72.64398084\n",
      "    30.77145      15.8791781    44.43410389]]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([  10.,    9.,  146.,  108.,   62.,   42.,   62.,   18.,    8.,\n",
       "         35.,   71.,   47.,   87.,   72.,   92.,   59.,   83.,   20.,\n",
       "        118.,   82.,   32.,   19.,   60.])"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "checkout_prop_res = []\n",
    "for t in range(300, 305):\n",
    "    train = start_out_df.drop('proportion',axis=1).head(t - 1)\n",
    "    train_X = train.drop('entiretraffic',axis=1)\n",
    "    train_y = train['entiretraffic']\n",
    "    test = start_out_df.drop('proportion',axis=1).loc[[t]]\n",
    "    test_X = test.drop('entiretraffic',axis=1)\n",
    "    test_y = test['entiretraffic']\n",
    "    dtrain = xgb.DMatrix(train_X, train_y)\n",
    "    dtest = xgb.DMatrix(test_X)\n",
    "    model = xgb.train(xgb_params, dtrain, num_boost_round=1000)\n",
    "    Et = model.predict(dtest)[0]\n",
    "    Pt = predict_proportion_ex(start_out_df, t)\n",
    "    checkout_prop_res.append(Et*Pt)\n",
    "print(checkout_prop_res[2])\n",
    "start_out_df.loc[302, 'proportion']*start_out_df.loc[302, 'entiretraffic']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Cluster Check-in Inference\n",
    "### HP-MSI"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "tmp_df = nyc_df[['stoptime','end station id']]\n",
    "tmp_df = tmp_df.merge(station[['id', 'bcluster']],\n",
    "            how='left',left_on='end station id',right_on='id').drop('id',axis=1)\n",
    "tmp_df['sdt'] = tmp_df['stoptime'].map(lambda x:datetime.strptime(x, '%m/%d/%Y %H:%M'))\n",
    "tmp_df['year'] = tmp_df['sdt'].map(lambda x:x.year)\n",
    "tmp_df['month'] = tmp_df['sdt'].map(lambda x:x.month)\n",
    "tmp_df['day'] = tmp_df['sdt'].map(lambda x:x.day)\n",
    "tmp_df['hour'] = tmp_df['sdt'].map(lambda x:x.hour)\n",
    "\n",
    "by_ymdh = dict(tmp_df.groupby(['year','month','day','hour']).groups)\n",
    "cnt_dict = {}\n",
    "for key in by_ymdh.keys():\n",
    "    ndict = dict(tmp_df.loc[by_ymdh[key]].groupby('bcluster').size())\n",
    "    count = np.array(list(map(lambda x:ndict[x] if x in ndict else 0, range(K1))))\n",
    "    cnt_dict[key] = count\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "series = pd.Series(cnt_dict)\n",
    "stop_in_df = pd.DataFrame({'count': series}).reset_index()\n",
    "first_id = list(map(lambda x:x[0], by_ymdh.values()))\n",
    "stop_in_df['weekday'] = list(tmp_df.loc[first_id, 'sdt'].map(lambda x:x.weekday()))\n",
    "stop_in_df['isweekend'] = stop_in_df['weekday'].map(lambda x:x>4)\n",
    "stop_in_df['proportion'] = stop_in_df['count'].map(lambda x:x/x.sum())\n",
    "stop_in_df['entiretraffic'] = stop_in_df['count'].map(lambda x:x.sum())\n",
    "stop_in_df = stop_in_df.drop('count', axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>year</th>\n",
       "      <th>month</th>\n",
       "      <th>day</th>\n",
       "      <th>hour</th>\n",
       "      <th>weekday</th>\n",
       "      <th>isweekend</th>\n",
       "      <th>proportion</th>\n",
       "      <th>entiretraffic</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>False</td>\n",
       "      <td>[0.0265486725664, 0.00884955752212, 0.07079646...</td>\n",
       "      <td>339</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>False</td>\n",
       "      <td>[0.00985221674877, 0.0197044334975, 0.08374384...</td>\n",
       "      <td>203</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>4</td>\n",
       "      <td>False</td>\n",
       "      <td>[0.00806451612903, 0.0, 0.104838709677, 0.0806...</td>\n",
       "      <td>124</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>4</td>\n",
       "      <td>False</td>\n",
       "      <td>[0.0694444444444, 0.0, 0.0694444444444, 0.0555...</td>\n",
       "      <td>72</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2014</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>4</td>\n",
       "      <td>False</td>\n",
       "      <td>[0.0508474576271, 0.0169491525424, 0.0, 0.0169...</td>\n",
       "      <td>59</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   year  month  day  hour  weekday  isweekend  \\\n",
       "0  2014      8    1     0        4      False   \n",
       "1  2014      8    1     1        4      False   \n",
       "2  2014      8    1     2        4      False   \n",
       "3  2014      8    1     3        4      False   \n",
       "4  2014      8    1     4        4      False   \n",
       "\n",
       "                                          proportion  entiretraffic  \n",
       "0  [0.0265486725664, 0.00884955752212, 0.07079646...            339  \n",
       "1  [0.00985221674877, 0.0197044334975, 0.08374384...            203  \n",
       "2  [0.00806451612903, 0.0, 0.104838709677, 0.0806...            124  \n",
       "3  [0.0694444444444, 0.0, 0.0694444444444, 0.0555...             72  \n",
       "4  [0.0508474576271, 0.0169491525424, 0.0, 0.0169...             59  "
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "stop_in_df = stop_in_df.rename(columns={'level_0':'year', 'level_1':'month', 'level_2':'day', 'level_3':'hour'})\n",
    "stop_in_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# params = {\n",
    "#     'rou': [0.7, 0.8],\n",
    "#     'alpha': [0.6, 0.2, 0.05, 0.3, 0.1, 0.8],\n",
    "#     'sigma': [stop_in_df['Temperature ° F'].std(), stop_in_df['Wind Speed mph'].std()],\n",
    "#     'H': 48,\n",
    "#     'J': 4,\n",
    "#     'q': 1.2\n",
    "# }"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[   5.1692893    10.06692452  121.16079405   82.40205826   39.14778581\n",
      "    39.17341633   51.17696141   20.3446158     4.72073223   39.80613841\n",
      "    65.17211469   45.60331195   79.19165691   70.17181633   52.26864596\n",
      "    51.07789259   57.25728519    5.87073233  126.54229964   73.1445516\n",
      "    30.98348805   15.98859738   44.74028774]]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([   7.,   10.,  135.,  106.,   61.,   59.,   62.,   22.,    5.,\n",
       "         39.,   86.,   33.,   85.,   73.,   96.,   54.,   47.,   19.,\n",
       "        142.,   72.,   43.,   23.,   58.])"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "checkin_prop_res = []\n",
    "for t in range(300, 305):\n",
    "    train = stop_in_df.drop('proportion',axis=1).head(t - 1)\n",
    "    train_X = train.drop('entiretraffic',axis=1)\n",
    "    train_y = train['entiretraffic']\n",
    "    test = stop_in_df.drop('proportion',axis=1).loc[[t]]\n",
    "    test_X = test.drop('entiretraffic',axis=1)\n",
    "    test_y = test['entiretraffic']\n",
    "    dtrain = xgb.DMatrix(train_X, train_y)\n",
    "    dtest = xgb.DMatrix(test_X)\n",
    "    model = xgb.train(xgb_params, dtrain, num_boost_round=1000)\n",
    "    Et = model.predict(dtest)[0]\n",
    "    Pt = predict_proportion_ex(start_out_df, t)\n",
    "    checkin_prop_res.append(Et*Pt)\n",
    "print(checkin_prop_res[2])\n",
    "stop_in_df.loc[302, 'proportion']*stop_in_df.loc[302, 'entiretraffic']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### P-TD"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "$$\n",
    "p(x)\n",
    "=\\frac{1}{\\sqrt{2\\pi}\\sigma x}\n",
    "\\mathrm{e}^{-\\frac{(\\ln x-\\mu)^2}{2\\sigma^2}}\n",
    "\\quad (x > 0)\n",
    "$$"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "$$\n",
    "\\begin{aligned}\n",
    "\\int_a^bp(x)\\mathrm{d}x\n",
    "=&\\frac{1}{\\sqrt{2\\pi}\\sigma}\\int_a^b\n",
    "\\mathrm{e}^{-\\frac{(\\ln x-\\mu)^2}{2\\sigma^2}}\\mathrm{d}\\ln x\\\\\n",
    "=&\\frac{1}{\\sqrt{2\\pi}\\sigma}\\int_\\alpha^\\beta\n",
    "\\mathrm{e}^{-\\frac{(y-\\mu)^2}{2\\sigma^2}}\\mathrm{d}y\\\\\n",
    "=&\\Phi_{\\mu,\\sigma}(\\beta)-\\Phi_{\\mu,\\sigma}(\\alpha)\\\\\n",
    "=&\\Phi_{\\mu,\\sigma}(\\ln b)-\\Phi_{\\mu,\\sigma}(\\ln a)\\\\\n",
    "=&\\Phi_{0,1}(\\frac{\\ln b-\\mu}{\\sigma})-\\Phi_{0,1}(\\frac{\\ln a-\\mu}{\\sigma})\n",
    "\\end{aligned}\n",
    "$$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>starthour</th>\n",
       "      <th>stophour</th>\n",
       "      <th>bcluster</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>995</th>\n",
       "      <td>5</td>\n",
       "      <td>5</td>\n",
       "      <td>13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>996</th>\n",
       "      <td>5</td>\n",
       "      <td>5</td>\n",
       "      <td>13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>997</th>\n",
       "      <td>5</td>\n",
       "      <td>5</td>\n",
       "      <td>13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>998</th>\n",
       "      <td>5</td>\n",
       "      <td>5</td>\n",
       "      <td>13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999</th>\n",
       "      <td>5</td>\n",
       "      <td>5</td>\n",
       "      <td>14</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     starthour  stophour  bcluster\n",
       "995          5         5        13\n",
       "996          5         5        13\n",
       "997          5         5        13\n",
       "998          5         5        13\n",
       "999          5         5        14"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "online_df = nyc_df[['starttime','stoptime','start station id','end station id']]\n",
    "online_df = online_df.merge(station[['id', 'bcluster']],\n",
    "            how='left',left_on='start station id',right_on='id').drop('id',axis=1)\n",
    "online_df['startdt'] = online_df['starttime'].map(lambda x:datetime.strptime(x, '%m/%d/%Y %H:%M'))\n",
    "online_df['stopdt'] = online_df['stoptime'].map(lambda x:datetime.strptime(x, '%m/%d/%Y %H:%M'))\n",
    "online_df['starthour'] = online_df['startdt'].map(lambda x:(x.day-1)*24+x.hour)\n",
    "online_df['stophour'] = online_df['startdt'].map(lambda x:(x.day-1)*24+x.hour)\n",
    "online_df = online_df[['starthour','stophour','bcluster']]\n",
    "online_df.head(1000).tail()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [],
   "source": [
    "def D_int(a, b, i, j):\n",
    "    norm = scipy.stats.norm(Mu[i, j],Sigma[i, j])\n",
    "    if a:\n",
    "        return norm.cdf(np.log(b*60)) - norm.cdf(np.log(a*60))\n",
    "    if b:\n",
    "        return norm.cdf(np.log(b*60))\n",
    "    return 0.0\n",
    "\n",
    "def p_td(t):\n",
    "    active = online_df[(online_df.starthour<=t) & (online_df.stophour>=t)]\n",
    "    longtrip = active[active.starthour < t]\n",
    "    shorttrip = active[active.starthour == t]\n",
    "    O = shorttrip.groupby('bcluster').size()\n",
    "    E = np.zeros(K1)\n",
    "    for i in range(K1):\n",
    "        T = [\n",
    "            tranmat_df.loc[t,'translation matrix'][j][i]\n",
    "            for j in longtrip['bcluster']\n",
    "        ]\n",
    "        D = [\n",
    "            D_int((t-x)*60, (t-x+1)*60, i, j)\n",
    "            for x,j in zip(longtrip['starthour'], longtrip['bcluster'])\n",
    "        ]\n",
    "        e1 = (np.array(T) * np.array(D)).sum()\n",
    "        e2 = 0\n",
    "        for dt in range(1,61):\n",
    "            for j in range(K1):\n",
    "                try:\n",
    "                    e2 += O[j]/60*tranmat_df.loc[t,'translation matrix'][j][i]*D_int(0,60-dt,i,j)\n",
    "                except:\n",
    "                    pass\n",
    "        E[i] = e1 + e2\n",
    "    return E"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\users\\小柒\\appdata\\local\\programs\\python\\python36\\lib\\site-packages\\ipykernel_launcher.py:5: RuntimeWarning: divide by zero encountered in log\n",
      "  \"\"\"\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([   6.27996446,    8.74958601,   99.1546424 ,   81.23222782,\n",
       "         54.22515036,   42.73701877,   51.66018427,   17.51216952,\n",
       "          2.03639041,   27.60991022,   64.6052096 ,   29.82814169,\n",
       "         62.76271033,   55.12903588,   79.10160808,   38.76790945,\n",
       "         43.97198888,   12.32862373,  111.43283155,   57.99334131,\n",
       "         28.70981207,   19.55083554,   42.13727825])"
      ]
     },
     "execution_count": 63,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "t = 302\n",
    "p_td(t)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([   7.,   10.,  135.,  106.,   61.,   59.,   62.,   22.,    5.,\n",
       "         39.,   86.,   33.,   85.,   73.,   96.,   54.,   47.,   19.,\n",
       "        142.,   72.,   43.,   23.,   58.])"
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "stop_in_df.loc[t, 'proportion']*stop_in_df.loc[t,'entiretraffic']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}