TeamAK-BaselineV1.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "%matplotlib inline\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "import os\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "from scipy.sparse import csr_matrix, hstack\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.cross_validation import StratifiedKFold\n",
    "from sklearn.metrics import log_loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "datadir = '..'\n",
    "gatrain = pd.read_csv(os.path.join(datadir,'gender_age_train.csv'),\n",
    "                      index_col='device_id')\n",
    "gatest = pd.read_csv(os.path.join(datadir,'gender_age_test.csv'),\n",
    "                     index_col = 'device_id')\n",
    "phone = pd.read_csv(os.path.join(datadir,'phone_brand_device_model.csv'))\n",
    "# Get rid of duplicate device ids in phone\n",
    "phone = phone.drop_duplicates('device_id',keep='first').set_index('device_id')\n",
    "events = pd.read_csv(os.path.join(datadir,'events_subset.csv'),\n",
    "                     parse_dates=['timestamp'], index_col='event_id')\n",
    "appevents = pd.read_csv(os.path.join(datadir,'app_events_subset.csv'), \n",
    "                        usecols=['event_id','app_id','is_active'],\n",
    "                        dtype={'is_active':bool})\n",
    "applabels = pd.read_csv(os.path.join(datadir,'app_labels_subset.csv'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "gatrain['trainrow'] = np.arange(gatrain.shape[0])\n",
    "gatest['testrow'] = np.arange(gatest.shape[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>phone_brand</th>\n",
       "      <th>device_model</th>\n",
       "      <th>brand</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>device_id</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>-8890648629457979026</th>\n",
       "      <td>小米</td>\n",
       "      <td>红米</td>\n",
       "      <td>51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1277779817574759137</th>\n",
       "      <td>小米</td>\n",
       "      <td>MI 2</td>\n",
       "      <td>51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5137427614288105724</th>\n",
       "      <td>三星</td>\n",
       "      <td>Galaxy S4</td>\n",
       "      <td>15</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3669464369358936369</th>\n",
       "      <td>SUGAR</td>\n",
       "      <td>时尚手机</td>\n",
       "      <td>9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>-5019277647504317457</th>\n",
       "      <td>三星</td>\n",
       "      <td>Galaxy Note 2</td>\n",
       "      <td>15</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                     phone_brand   device_model  brand\n",
       "device_id                                             \n",
       "-8890648629457979026          小米             红米     51\n",
       " 1277779817574759137          小米           MI 2     51\n",
       " 5137427614288105724          三星      Galaxy S4     15\n",
       " 3669464369358936369       SUGAR           时尚手机      9\n",
       "-5019277647504317457          三星  Galaxy Note 2     15"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "phone.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Brand features: train shape (74645, 131), test shape (112071, 131)\n"
     ]
    }
   ],
   "source": [
    "brandencoder = LabelEncoder().fit(phone.phone_brand)\n",
    "phone['brand'] = brandencoder.transform(phone['phone_brand'])\n",
    "gatrain['brand'] = phone['brand']\n",
    "gatest['brand'] = phone['brand']\n",
    "Xtr_brand = csr_matrix((np.ones(gatrain.shape[0]), \n",
    "                       (gatrain.trainrow, gatrain.brand)))\n",
    "Xte_brand = csr_matrix((np.ones(gatest.shape[0]), \n",
    "                       (gatest.testrow, gatest.brand)))\n",
    "print('Brand features: train shape {}, test shape {}'.format(Xtr_brand.shape, Xte_brand.shape))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model features: train shape (74645, 1667), test shape (112071, 1667)\n"
     ]
    }
   ],
   "source": [
    "m = phone.phone_brand.str.cat(phone.device_model)\n",
    "modelencoder = LabelEncoder().fit(m)\n",
    "phone['model'] = modelencoder.transform(m)\n",
    "gatrain['model'] = phone['model']\n",
    "gatest['model'] = phone['model']\n",
    "Xtr_model = csr_matrix((np.ones(gatrain.shape[0]), \n",
    "                       (gatrain.trainrow, gatrain.model)))\n",
    "Xte_model = csr_matrix((np.ones(gatest.shape[0]), \n",
    "                       (gatest.testrow, gatest.model)))\n",
    "print('Model features: train shape {}, test shape {}'.format(Xtr_model.shape, Xte_model.shape))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "130 1666\n"
     ]
    }
   ],
   "source": [
    "print(np.max(phone.brand),np.max(phone.model))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>device_id</th>\n",
       "      <th>app</th>\n",
       "      <th>size</th>\n",
       "      <th>trainrow</th>\n",
       "      <th>testrow</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>-9220329415676028483</td>\n",
       "      <td>61</td>\n",
       "      <td>1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>83094</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>-9220329415676028483</td>\n",
       "      <td>166</td>\n",
       "      <td>1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>83094</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>-9220329415676028483</td>\n",
       "      <td>183</td>\n",
       "      <td>1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>83094</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>-9220329415676028483</td>\n",
       "      <td>476</td>\n",
       "      <td>1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>83094</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>-9220329415676028483</td>\n",
       "      <td>615</td>\n",
       "      <td>1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>83094</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "             device_id  app  size  trainrow  testrow\n",
       "0 -9220329415676028483   61     1       NaN    83094\n",
       "1 -9220329415676028483  166     1       NaN    83094\n",
       "2 -9220329415676028483  183     1       NaN    83094\n",
       "3 -9220329415676028483  476     1       NaN    83094\n",
       "4 -9220329415676028483  615     1       NaN    83094"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "appencoder = LabelEncoder().fit(appevents.app_id)\n",
    "appevents['app'] = appencoder.transform(appevents.app_id)\n",
    "napps = len(appencoder.classes_)\n",
    "deviceapps = (appevents.merge(events[['device_id']], how='left',left_on='event_id',right_index=True)\n",
    "                       .groupby(['device_id','app'])['app'].agg(['size'])\n",
    "                       .merge(gatrain[['trainrow']], how='left', left_index=True, right_index=True)\n",
    "                       .merge(gatest[['testrow']], how='left', left_index=True, right_index=True)\n",
    "                       .reset_index())\n",
    "deviceapps.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Apps data: train shape (74645, 5504), test shape (112071, 5504)\n"
     ]
    }
   ],
   "source": [
    "d = deviceapps.dropna(subset=['trainrow'])\n",
    "Xtr_app = csr_matrix((np.ones(d.shape[0]), (d.trainrow, d.app)), \n",
    "                      shape=(gatrain.shape[0],napps))\n",
    "d = deviceapps.dropna(subset=['testrow'])\n",
    "Xte_app = csr_matrix((np.ones(d.shape[0]), (d.testrow, d.app)), \n",
    "                      shape=(gatest.shape[0],napps))\n",
    "print('Apps data: train shape {}, test shape {}'.format(Xtr_app.shape, Xte_app.shape))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "applabels = applabels.loc[applabels.app_id.isin(appevents.app_id.unique())]\n",
    "applabels['app'] = appencoder.transform(applabels.app_id)\n",
    "labelencoder = LabelEncoder().fit(applabels.label_id)\n",
    "applabels['label'] = labelencoder.transform(applabels.label_id)\n",
    "nlabels = len(labelencoder.classes_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>device_id</th>\n",
       "      <th>label</th>\n",
       "      <th>size</th>\n",
       "      <th>trainrow</th>\n",
       "      <th>testrow</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>-9220329415676028483</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>83094</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>-9220329415676028483</td>\n",
       "      <td>12</td>\n",
       "      <td>8</td>\n",
       "      <td>NaN</td>\n",
       "      <td>83094</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>-9220329415676028483</td>\n",
       "      <td>13</td>\n",
       "      <td>31</td>\n",
       "      <td>NaN</td>\n",
       "      <td>83094</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>-9220329415676028483</td>\n",
       "      <td>20</td>\n",
       "      <td>3</td>\n",
       "      <td>NaN</td>\n",
       "      <td>83094</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>-9220329415676028483</td>\n",
       "      <td>22</td>\n",
       "      <td>3</td>\n",
       "      <td>NaN</td>\n",
       "      <td>83094</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "             device_id  label  size  trainrow  testrow\n",
       "0 -9220329415676028483      0     1       NaN    83094\n",
       "1 -9220329415676028483     12     8       NaN    83094\n",
       "2 -9220329415676028483     13    31       NaN    83094\n",
       "3 -9220329415676028483     20     3       NaN    83094\n",
       "4 -9220329415676028483     22     3       NaN    83094"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "devicelabels = (deviceapps[['device_id','app']]\n",
    "                .merge(applabels[['app','label']])\n",
    "                .groupby(['device_id','label'])['app'].agg(['size'])\n",
    "                .merge(gatrain[['trainrow']], how='left', left_index=True, right_index=True)\n",
    "                .merge(gatest[['testrow']], how='left', left_index=True, right_index=True)\n",
    "                .reset_index())\n",
    "devicelabels.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Labels data: train shape (74645, 30), test shape (112071, 30)\n"
     ]
    }
   ],
   "source": [
    "d = devicelabels.dropna(subset=['trainrow'])\n",
    "Xtr_label = csr_matrix((np.ones(d.shape[0]), (d.trainrow, d.label)), \n",
    "                      shape=(gatrain.shape[0],nlabels))\n",
    "d = devicelabels.dropna(subset=['testrow'])\n",
    "Xte_label = csr_matrix((np.ones(d.shape[0]), (d.testrow, d.label)), \n",
    "                      shape=(gatest.shape[0],nlabels))\n",
    "print('Labels data: train shape {}, test shape {}'.format(Xtr_label.shape, Xte_label.shape))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "All features: train shape (74645, 7332), test shape (112071, 7332)\n"
     ]
    }
   ],
   "source": [
    "Xtrain = hstack((Xtr_brand, Xtr_model, Xtr_app, Xtr_label), format='csr')\n",
    "Xtest =  hstack((Xte_brand, Xte_model, Xte_app, Xte_label), format='csr')\n",
    "print('All features: train shape {}, test shape {}'.format(Xtrain.shape, Xtest.shape))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "targetencoder = LabelEncoder().fit(gatrain.group)\n",
    "y = targetencoder.transform(gatrain.group)\n",
    "nclasses = len(targetencoder.classes_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def score(clf, random_state = 0):\n",
    "    kf = StratifiedKFold(y, n_folds=5, shuffle=True, random_state=random_state)\n",
    "    pred = np.zeros((y.shape[0],nclasses))\n",
    "    for itrain, itest in kf:\n",
    "        Xtr, Xte = Xtrain[itrain, :], Xtrain[itest, :]\n",
    "        ytr, yte = y[itrain], y[itest]\n",
    "        clf.fit(Xtr, ytr)\n",
    "        pred[itest,:] = clf.predict_proba(Xte)\n",
    "        # Downsize to one fold only for kernels\n",
    "        return log_loss(yte, pred[itest, :])\n",
    "        print(\"{:.5f}\".format(log_loss(yte, pred[itest,:])), end=' ')\n",
    "    print('')\n",
    "    return log_loss(y, pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfUAAAFZCAYAAACBlraEAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzt3Xl8lPW5///XZCM7kA0IYYcQEgRZhFDFiCRQF0CsVBRw\nqRb1fF0AbcV6XPp7IOZAbaVyejBApYoaRULZjQSroqxhT9jDlkCABLKQkH3m9wcSRZYMyST3LO/n\nXyfDLNft+fR+577mk/syWSwWCyIiIuLw3IwuQERERGxDoS4iIuIkFOoiIiJOQqEuIiLiJBTqIiIi\nTkKhLiIi4iQ8rHnSjBkz2LZtGzU1NUycOJGEhIQrnvPOO++wY8cOPvrooyte89RTTxEfH88rr7xC\nRkYGLVu2BOCJJ54gLi7OhocjIiLiuuoM9U2bNpGVlUVycjKFhYWMHj36ilDPysoiPT0dT0/Pa74m\nPj4egJdeeklBLiIi0gjqDPUBAwbQu3dvAAIDAykrK8NisWAymWqfk5iYyJQpU3jvvfeu+xoRERFp\nPHV+p24ymfD29gZg0aJFxMXFXRboS5YsYeDAgYSHh1v1moULF/Loo4/y4osvUlhYaNODERERcWVW\nb5RLS0sjJSWF1157rfaxoqIiUlJSePzxx7FYLFdcjf/yNaNGjeLFF1/kX//6F927d6+9shcREZGG\nsyrU161bR1JSEvPmzcPf37/28Y0bN1JQUMC4ceN47rnn2Lt3L4mJidd8TWxsLFFRUQAMHTqUAwcO\nXPdz1bIXERGxnqmugS4lJSU8/PDDLFiwgKCgoGs+78SJE7zyyit8+OGHnD9/nnHjxl3xmueff54/\n/OEPtGvXjk8++YSsrKzLrvyvJi/v/A0eksi1hYYGaE2JTWlNia2FhgbU+7V1bpRbtWoVhYWFTJo0\nqXaDXGxsLJGRkbU72n9p9erVV7xmxowZjBs3jsmTJ+Pj44Ofnx/Tp0+vd+EiIiJyuTqv1I2m34DF\nlnRVJbamNSW21pArdd1RTkRExEko1EVERJyEQl1ERMRJKNRFRESchEJdRETESSjURUREnIRCXURE\nxEko1EVERJyEQl1ERMRJKNRFRESchEJdRETESSjURUREnIRCXURExEko1EVERJyEQl1ERMRJeBhd\nwPWM+sMywoN9uWdQRwZGtzK6HBEREbtm11fqZrOFnLxS3l+WyaY9p40uR0RExK7Zdaj/3MoNx4wu\nQURExK45TKjnni01ugQRERG75jCh3jrI1+gSRERE7JrDhLqPtwcWi8XoMkREROyWXYe6u5uJiFA/\nwlr6cCiniPUZp4wuSURExG7Z9Z+0/XvmSPLyzpNXWMabH2zmo6/20zk8kDbBfkaXJiIiYnfs+kr9\nktAWPjx2Vw8qq8zMWZpJVXWN0SWJiIjYHYcIdYBbosKIuzmc7DMlfP51ltHliIiI2B2HCXWAsUO7\n0TbEj7Xbcth+IM/ockREROyKQ4V6M093nh4Vg6eHG/9ctZdzxeVGlyQiImI3HCrUAdqG+vNQfDdK\ny6t5f1kmNWaz0SWJiIjYBYcLdYC43uH0jwrjYE4Ry74/anQ5IiIidsEhQ91kMvHYr7sT0tybFeuP\nsvdYgdEliYiIGM4hQx3A19uTp0bG4OZmIml5JsUXKo0uSURExFAOG+oAXdo25/7bO1NUUsk/V+7F\nrNvIioiIC3PoUAcYPrA9MZ2C2JV1lrQt2UaXIyIiYhiHD3U3k4kn740m0M+LRd9kcSS32OiSRERE\nDOHwoQ7Q3M+L398bTY3ZwvtLMymrqDa6JBERkSZnVajPmDGDsWPHMmbMGNasWXPV57zzzjtMmDDh\nuq85deoUEyZMYPz48UyePJmqqiobHMJFMZ2CuDu2A2cKy/gwdb/GtIqIiMupM9Q3bdpEVlYWycnJ\nzJ07l+nTp1/xnKysLNLT0zGZTNd9zaxZs5gwYQILFy6kffv2LF682KYHc9/gTnQJD2TTntN8vzvX\npu8tIiJi7+oM9QEDBjBr1iwAAgMDKSsru+IqODExkSlTplz3NWazmc2bNzNkyBAAhgwZwvr16212\nIAAe7m48NTIGn2YefLzmACfzS236/iIiIvaszlA3mUx4e3sDsGjRIuLi4mqvyAGWLFnCwIEDCQ8P\nv+Zr7rjjDtzc3CgrK8PT0xOA4OBg8vJsP5QlpIUPj98VVTumtbJKY1pFRMQ1eFj7xLS0NFJSUpg/\nf37tY0VFRaSkpLBgwQJyc3OvuIK/9Jp//vOfAJf9MmDtd96hoQHWlljrrtAAjpwuYfWGoyzbcIxn\nftP7ht9DnFd91pTI9WhNib2wKtTXrVtHUlIS8+fPx9/fv/bxjRs3UlBQwLhx46ioqCA7O5vExESm\nTp162Wv8/PwA8PX1pbKyEi8vL06fPk1YWFidn52Xd75eBzbqVx3YdSiPVeuP0qlVAP26h9brfcS5\nhIYG1HtNiVyN1pTYWkN+Sayz/V5SUsLMmTOZM2cOAQGXf9Dw4cNZsWIFycnJzJ49m+joaKZOncr5\n8+ev+ppBgwaRmpoKQGpqKoMHD6534XXx8nTn6VE98fJw44NVe8kvKmu0zxIREbEHdV6pr1q1isLC\nQiZNmoTFYsFkMhEbG0tkZCTx8fFXfc3q1auveM2MGTN47rnnePnll/nss88IDw9n9OjRNj+gn2sb\n4sfDCZEsWL2PpGV7eHlcH9zdnOJP80VERK5gstj5H3Q3tK1lsVh4f1kmm/ee4d5fdeD+27vYqDJx\nRGqViq1pTYmtNWr73dGZTCYeGR5FSHNvVq4/xp6j54wuSUREpFE4fagD+Hp78PSonri5mZi7fA/F\npRrTKiIizsclQh2gc3ggv4nrQlFpJfNW7tGYVhERcTouE+oAwwa0o2fnIDIOn+OrzRrTKiIizsWl\nQt3NZOLJe6Jp7ufF4m81plVERJyLS4U6QKCfF0+OiMZstjBnaQYXyjWmVUREnIPLhTpATMcg7h7U\ngbzCcj5M3acxrSIi4hRcMtQBRt3Wia5tm7N57xnW7dKYVhERcXwuG+oe7m5MHBmNbzMPPllzgBMa\n0yoiIg7OZUMdIKS5D4/fHUVltZk5SzM0plVERByaS4c6QL/uYQzp25YTeaUkf33I6HJERETqzeVD\nHWDsnV2JCPXnm+0nSN93xuhyRERE6kWhDnh6uPP0qBi8PN34YPU+8gs1plVERByPQv1H4SF+jIuP\npKyimveXZ1JdYza6JBERkRuiUP+Z23q1YWB0K7JOFLP0+yNGlyMiInJDFOo/c3FMa3dCW3izasMx\nMo9oTKuIiDgOhfov+DT72ZjWFXso0phWERFxEAr1q+jUJpAH7uhCcWkl81ZoTKuIiDgGhfo1JNzS\njl5dgsk8co7UzceNLkdERKROCvVrcDOZ+N09PWju70XKt4fJOllkdEkiIiLXpVC/jkBfLybee3FM\n6/tLM7lQXmV0SSIiItekUK9Dj45B3POrjuQXlbPgy/0a0yoiInZLoW6FUbd1pGtEc9L3neG7nSeN\nLkdEROSqFOpWcHdz46kRMfh5e/BJ2kFO5JUYXZKIiMgVFOpWCm7uzeN396Cq2sycpZlUaEyriIjY\nGYX6DegbGcrQvhGcyC/ls7UHjS5HRETkMgr1G/TbO7vQLsyfb3acZIvGtIqIiB1RqN+gn49pXbB6\nH3ka0yoiInZCoV4PbYL9GJ/Q/eKY1mUa0yoiIvZBoV5Pt97UmtiYVhw+WcySdYeNLkdEREShXl8m\nk4kJw7oT1tKH1RuPk3HkrNEliYiIi1OoN8DFMa0xuLuZmLd8D0UlFUaXJCIiLkyh3kAdWwcyZkhX\nii9UMVdjWkVExEAKdRtI6B9B7y7B7DlawJebNKZVRESMoVC3AdOPY1pb/Dim9dAJjWkVEZGmZ1Wo\nz5gxg7FjxzJmzBjWrFlz1ee88847TJgwofbnAwcOkJCQwMcff1z72CuvvMKIESN45JFHeOSRR/j2\n228bWL79CPD1YuKIGCwWjWkVERFjeNT1hE2bNpGVlUVycjKFhYWMHj2ahISEy56TlZVFeno6np6e\nAJSVlTFt2jQGDRp0xfu99NJLxMXF2ah8+xLVoSUjbu3Ish+O8sHqffzXfT0xmUxGlyUiIi6iziv1\nAQMGMGvWLAACAwMpKyu7YqZ4YmIiU6ZMqf3Zy8uLefPmERYWZuNy7d+IWzsSGdGcrfvz+HaHxrSK\niEjTqTPUTSYT3t7eACxatIi4uLjLrj6XLFnCwIEDCQ8Pr33M3d0dLy+vq77fwoULefTRR3nxxRcp\nLCxsaP12x93NjYkjL45p/XTtQXLOaEyriIg0jTrb75ekpaWRkpLC/Pnzax8rKioiJSWFBQsWkJub\ne8UV/C+NGjWKFi1aEBUVRVJSEu+99x6vvfbadV8TGhpgbYl2IzQ0gMkP9WXaB5uZu3IPf50Uh7eX\n1f+ppZE54poS+6Y1JfbCqqRZt24dSUlJzJ8/H39//9rHN27cSEFBAePGjaOiooLs7GwSExOZOnXq\nVd8nNja29v8eOnQob775Zp2fnZd33poS7U7nVv7E94sgbWsO7yVv57G7oowuSbh48nXUNSX2SWtK\nbK0hvyTW2X4vKSlh5syZzJkzh4CAyz9o+PDhrFixguTkZGbPnk10dPQ1Ax3g+eefJzs7G7i4AS8y\nMrLehTuCMUO60r6VP9/tPMnmvaeNLkdERJxcnVfqq1atorCwkEmTJmGxWDCZTMTGxhIZGUl8fPxV\nX5OZmUliYiInT57Ew8OD1NRUZs+ezbhx45g8eTI+Pj74+fkxffp0mx+QPfH0cOPpUT358wdb+NeX\n++jYJpCwFj5GlyUiIk7KZKnri3CDOUNb64fducxfuZdObQJ5ZXxfPNx1zx+jqFUqtqY1JbbWqO13\nabhbb2rDoJjWHMktJuU7jWkVEZHGoVBvIuOHRdKqpQ9fbjrO7sMa0yoiIranUG8iF8e09sTD3cS8\nFXso1JhWERGxMYV6E+rQOoAxQ7py/kIVc5drTKuIiNiWQr2JxfeL4OauIew9VsDqjceMLkdERJyI\nQr2JXRrT2jKgGUu+O8KhHI1pFRER21CoG8Dfx5OJI6KxYOH9ZRmUakyriIjYgELdIN3bt2TkrZ04\nW1zBglX76rxvvoiISF0U6gYa8auOdG/Xgq0H8vhm+wmjyxEREQenUDeQm5uJ34+Ixt/Hk0/XHiJb\nY1pFRKQBFOoGCwr05nf39KC6xsycpRlUVNYYXZKIiDgohboduLlrCAn925F79gKfpB0wuhwREXFQ\nCnU78cAdXejQKoB1u3LZuOeU0eWIiIgDUqjbiYtjWmNo5uXOh1/u50zBBaNLEhERB6NQtyOtgnx5\nZFh3yitrmLM0k+oas9EliYiIA1Go25lBPVtza8/WHD11nsXfZhldjoiIOBCFuh0aNyySVkG+pG7O\nZldWvtHliIiIg1Co2yFvLw+eGRXz45jWvRSc15hWERGpm0LdTrVvFcCDd3ajpKyKucszMZt1G1kR\nEbk+hbodu7NvW/p0C2Hf8UJWakyriIjUQaFux0wmE4/ffXFM69J1RziQXWh0SSIiYscU6nbO38eT\np0bGYMFC0vJMSso0plVERK5Ooe4AItu1YNRtnThXXMEHq/ZqTKuIiFyVQt1B3DuoI1HtW7D9YD5f\nb9OYVhERuZJC3UFcHNMag7+PJ599fZDjp88bXZKIiNgZhboDaRnQjCfu6UF1jYU5SzM1plVERC6j\nUHcwvbuGMOyWdpw6d4GP12hMq4iI/ESh7oAeuKMLHVoH8P3uXDZkakyriIhcpFB3QB7uPxvTmrqf\n0xrTKiIiKNQdVquWvjw6vDsVlTXM+XcmVdUa0yoi4uoU6g4sNqY1t93UhmOnNaZVREQU6g5vXEIk\nrYN8+WpLNjsOaUyriIgrU6g7uGZe7jw9KgYPdzf+uVJjWkVEXJlC3Qm0bxXA2KFdKSmrImmZxrSK\niLgqhbqTGNKnLX0jQ9mfXciKDUeNLkdERAxgVajPmDGDsWPHMmbMGNasWXPV57zzzjtMmDCh9ucD\nBw6QkJDAxx9/XPvYqVOnmDBhAuPHj2fy5MlUVWnimK1cHNMaRXBgM5Z+rzGtIiKuqM5Q37RpE1lZ\nWSQnJzN37lymT59+xXOysrJIT0/HZDIBUFZWxrRp0xg0aNBlz5s1axYTJkxg4cKFtG/fnsWLF9vo\nMATAz9uTiSNjMGHi/WUa0yoi4mrqDPUBAwYwa9YsAAIDAykrK7ti9GdiYiJTpkyp/dnLy4t58+YR\nFhZ22fM2b97MkCFDABgyZAjr169v8AHI5bpFtGDU4E4UnK/gnys1plVExJXUGeomkwlvb28AFi1a\nRFxcXO0VOcCSJUsYOHAg4eHhtY+5u7vj5eV1xXuVl5fj6ekJQHBwMHl5eQ0+ALnSPbEd6NGhJTsO\n5bN2a47R5YiISBPxsPaJaWlppKSkMH/+/NrHioqKSElJYcGCBeTm5t7QVaG1zw0NDbD6PeUnLz96\nCy/89Rs+/08WA24Kp0tEC6NLshtaU2JrWlNiL6wK9XXr1pGUlMT8+fPx9/evfXzjxo0UFBQwbtw4\nKioqyM7OJjExkalTp171fXx9famsrMTLy4vTp09f0Z6/mrw8zQ2vr8fv6sG7i3by9oLNvPH4LXh7\nWf07nNMKDQ3QmhKb0poSW2vIL4l1tt9LSkqYOXMmc+bMISDg8g8aPnw4K1asIDk5mdmzZxMdHX3N\nQAcYNGgQqampAKSmpjJ48OB6Fy5169UlmF8PaM/pgjI+/kpjWkVEnF2dl26rVq2isLCQSZMmYbFY\nMJlMxMbGEhkZSXx8/FVfk5mZSWJiIidPnsTDw4PU1FRmz57Nc889x8svv8xnn31GeHg4o0ePtvkB\nyeXuj+vM/uwCfsg4RY+OLflVzzZGlyQiIo3EZLHz7dFqazXcmcIy3vznZiwWeOPxW2gd5Gt0SYZR\nq1RsTWtKbK1R2+/i+MJa+PDor6OoqKphztIMjWkVEXFSCnUXMTC6FYN7teH46RIWfXPI6HJERKQR\nKNRdyMPxkbQJ9iUtPYftB3WPABERZ6NQdyHNvNx5ZlTP2jGt54rLjS5JRERsSKHuYiLC/Hkovhul\n5dUkLd+jMa0iIk5Eoe6C7rg5nH7dQzmQXcjy9UeNLkdERGxEoe6CTCYTj90VRXCgN8t+OML+4wVG\nlyQiIjagUHdRft6ePDXq4pjWpOV7OH+h0uiSRESkgRTqLqxr2+aMvl1jWkVEnIVC3cXdFduB6I4t\n2Zl1lrR0jWkVEXFkCnUX52Yy8ft7own09eTz/xzi6Klio0sSEZF6UqgLzf2b8eS90dSYLcxZmklZ\nRbXRJYmISD0o1AWAnp2DuWtge84UlLFQY1pFRBySQl1qjb69M53DA9mQeYofducaXY6IiNwghbrU\n8nB346mRMfg0c2fhVwfIPVtqdEkiInIDFOpymdCfjWl9f2kmVdU1RpckIiJWUqjLFQb0aMXtvcM5\nfqaEz/+TZXQ5IiJiJYW6XNVD8d0ID/Fj7dYcth/QmFYREUegUJeraubpztOjYvD0cOOfqzSmVUTE\nESjU5ZoiQn82pnVZJjVms9EliYjIdSjU5brieofTPyqMAzlFLP/hqNHliIjIdSjU5bpMJhOP/bo7\nIc29Wf7DUfYe05hWEZHGsGnPaV6fv4lRf1hW7/dQqEudfL09eWpkDCaTibnLMynWmFYREZvatOc0\n7y/LJCevFLO5/hMzFepilS5tm3N/XGcKSyo1plVExMZWbjhqk/dRqIvVfj2wPTEdW7Ir6yxrtmQb\nXY6IiNM4mW+bO3gq1MVqbiYTT46IIdDPi0XfZHEkV2NaRUQa6vjp85hMJpu8l0JdbkhzPy9+/+OY\n1vc1plVEpEF2ZeXz9sfbqGnA9+g/p1CXGxbTKYi7YztwprCMj1L36/t1EZF6+HpbDrO+2IXZbOG/\n7uvJUyNjiAj1x92t/lftHjasT1zIfYM7sf94ARv3nCa6YxC39WpjdEkiIg7BbLbw+X8O8dWWbAJ9\nPXnugV50CW8OwMDoVoSGBtT7vXWlLvXy05hWDxau2a8xrSIiVqiorOF/l+zmqy3ZhIf48d+P9K8N\ndFtQqEu9hbTw4bG7oqisMvN//9aYVhGR6yksqSDxk21sP5hPjw4t+dP4voS08LHpZyjUpUFuiQrj\njpvDyckr4bOvDxldjoiIXcrJK+GtD9M5duo8t/Vqw+Tf9sbX29Pmn6Pv1KXBxg7txsETRXy97QQ9\nOgTRr3uo0SWJiNiNjCNn+b9/Z1BWUcNv4jpzd2wHm/0J2y/pSl0azMvTnadHxuDl4cYHq/Zytkhj\nWkVEAL7dcYJ3P99FVbWFp0fFcM+gjo0W6KBQFxtpG+rPwwmRXKio5v3lGtMqIq7NbLGw6JtD/OvL\n/fh6e/DHh/owoEerRv9cq9rvM2bMYNu2bdTU1DBx4kQSEhKueM4777zDjh07+OijjwB4++232blz\nJyaTiVdffZWePXvyyiuvkJGRQcuWLQF44okniIuLs+HhiJEG92rDnqPn2Lz3DEu/P8r9t3c2uiQR\nkSZXWVXDvBV7SN+fR6sgXyaP6UVYS98m+ew6Q33Tpk1kZWWRnJxMYWEho0ePviLUs7KySE9Px9Pz\n4pf+W7Zs4dixYyQnJ5OVlcWrr75KcnIyAC+99JKC3EmZTCYeGR7F4ZPFrFx/lB7tW9CjY5DRZYmI\nNJni0kr+vngXh08WE9muBc/efxP+PrbfEHctdbbfBwwYwKxZswAIDAykrKzsijuIJSYmMmXKlNqf\nN2zYQHx8PABdunShuLiY0lL9HbMr8PX24OlRPXFzM5G0fA/FpRrTKiKu4WR+KdM+TOfwyWIGxbTm\nxQdvbtJABytC3WQy4e3tDcCiRYuIi4u77Ev+JUuWMHDgQMLDw2sfy8/PJyjopyu0oKAg8vPzAVi4\ncCGPPvooL774IoWFhTY7ELEfncMDuT+uM0WllcxfuRezbiMrIk5u77ECpn+0lfyickbd1okn7+2B\np0fTb1uz+k/a0tLSSElJYf78+bWPFRUVkZKSwoIFC8jNzb3ma80/bpoaNWoULVq0ICoqiqSkJN57\n7z1ee+21635uQ26XJ8YZf3cMWSfPs23/GdbvOcPoO7oaXVItrSmxNa0p15a2+TizF+3AZIIpD/dl\nSL92htViVaivW7eOpKQk5s+fj7+/f+3jGzdupKCggHHjxlFRUUF2djaJiYmEhYWRl5dX+7wzZ84Q\nGhpKhw4dah8bOnQob775Zp2fnZd3/gYOR+zJI8Miycop5F8r99A2yIdObQKNLonQ0ACtKbEprSnX\nZbFYWLLuCCvWH8XP24Nn77+J7u1bNHg9NOq930tKSpg5cyZz5swhIODyDxo+fDgrVqwgOTmZ2bNn\nEx0dzdSpU7n11lv56quvAMjMzKRVq1b4+vry/PPPk52dDVzcgBcZGVnvwsX+Bfp58eSIaMxmC3OW\nZmhMq4g4jarqGpKW72HF+qOEtfDhTxP60b19S6PLqvtKfdWqVRQWFjJp0iQsFgsmk4nY2FgiIyNr\nN8P9Up8+fYiJiWHs2LG4u7vz+uuvAzBu3DgmT56Mj48Pfn5+TJ8+3bZHI3YnpmMQdw/qwMoNx/jX\nl/t4amRMo954QUSksZ2/UMnslN0czCmia0Rznrv/JgJ8vYwuCwCTxc6HYaut5fiqa8zM+GQ7h04U\n8fhdUQzuHV73ixqJWqVia1pTruX0uQv8bdFOzhSUMaBHGE/c0wNPD3ebfoZGr4pd83B3Y+LIaHyb\nefDxmgOczNefN4qI4zmQXci0D9M5U1DGvb/qwMSRMTYP9IZSqEuTCGn+45jWajNzlmZQWaUxrSLi\nODZknuIvydspr6zh8bujuP/2LrjZ4VeJCnVpMv2jwhjSpy05eaUa0yoiDsFisbDs+yPMXb4HTw93\nJv+2N4N7GfcVYl0U6tKkHryzKxGhfvxn+wnS950xuhwRkWuqrjEzf+Ve/v39EUKae/OnCf2ItvNb\nXyvUpUl5ebrz9KieeHm6sWD1PvKLyowuSUTkCiVlVfz1sx2szzhFpzaBvPpIf9qG+BldVp0U6tLk\nwkP8GBf/45jWZZlU12hMq4jYjzMFF5j+0Vb2HS+kX/dQ/vhwH5r72cefrNVFoS6GuK1XGwZGtyLr\nRDFLvz9idDkiIgAcyili2odbOXXuAr8e2J5n7utJM0/72uF+PQp1McTFMa3dCW3hzaoNx8g8es7o\nkkTExW3ee5oZn27nQnk1jwzvzm+HdLXLHe7Xo1AXw/g0+2lM67zleyjSmFYRMYDFYmHlhqPMWZqJ\nh7uJSWN6cUeftkaXVS8KdTFUpzaB/Cauy8UxrSv2aEyriDSp6hozC1bvY/G3hwkKbMafxvejZ+dg\no8uqN4W6GG7YgHbc1DmYjCPnSN183OhyRMRFXCiv4m+f72Tdrlw6tA7gvx/pT0SYf90vtGMKdTGc\nm8nEE/f0oLm/FynfHibrZJHRJYmIk8svLGP6wm3sPVbAzV1DmPpwX1r4NzO6rAZTqItdCPTzYuK9\nF8e0vr80kwvlGtMqIo3jSG4x0z7aysn8UhL6t+PZ+2+imZfj7HC/HoW62I0eHYO451cdyS8q519f\n7sPOBwiKiAPauj+P//l4G+cvVDIuIZKH4rvh5uZYO9yvR6EudmXUbR3pGtGcLfvOsG5XrtHliIiT\nsFgsfLnpOP9YshuTycTzv+nF0H4RRpdlcwp1sSvubm48NSIGP28PPllzgBN5JUaXJCIOrsZs5qOv\nDvD5fw7R3N+LqeP60rtriNFlNQqFutid4ObePHZXjx/HtGZqTKuI1FtZRTWzvtjFN9tP0C7Mn/9+\npD8dWgcYXVajUaiLXerXPZQ7+7blRH4pyWsPGl2OiDigc8XlvL1wGxmHz9GrSzBTx/UlKNDb6LIa\nlUJd7NbFMa3+fLPjJFs0plVEbsCxU+eZ9mE6OXklDOnblud+cxM+zTyMLqvRKdTFbnl6uPPMfTE/\njWkt1JhWEanbjoP5JH68jaKSSsbe2ZXxCZG4u7lG3LnGUYrDahPsx/iE7pRpTKuIWCEtPZv3UnZh\nsVj4f/fwfuglAAAcIUlEQVTfxLAB7TE52FCWhlCoi9279abWxMa0IutkMf9epzGtInIls9nCJ2sO\n8EnaQQJ8vXh5XF/6RoYaXVaTU6iL3TOZTEwY1p2wlj6s2niMjCNnjS5JROxIeWU1s1N2k7Y1h7Yh\nfvz3I/3o1CbQ6LIMoVAXh3BxTGsM7pfGtJZUGF2SiNiBgvMV/M/H29lxKJ+Yji15ZXw/Qpr7GF2W\nYRTq4jA6tg5kzB1dKL5QxTyNaRVxedlnSpj2YTrHTp/n9t7hvDCmN77ezr/D/XoU6uJQEm5pR68u\nwWQeLeDLTRrTKuKqdh8+y/SFWyk4X8GYO7rw6K+74+GuSNN/AXEoJpOJ393TgxaXxrSe0JhWEVfz\nn+0nmLVoFzU1Fp65ryd3xXZwqR3u16NQF4cT6OvFxBExWCwW3l+WyYXyKqNLEpEmYLZY+Ozrg3yU\nuh8/Hw/++HAfbokKM7osu6JQF4cU1aElI269OKZ1wZf7NaZVxMlVVNXwjyUZpG7Opk2wL68+0p+u\nbZsbXZbdUaiLwxpxa0ciI5qTvu8M3+48aXQ5ItJIikoqmPHJNrYdyCOqfQv+NKEfYS1cd4f79SjU\nxWG5u7kxceTFMa2fph0kR2NaRZzOibwSpn24lSO557n1ptZMefBm/Lw9jS7LbinUxaEFBXrzu7t7\nUPXjmNYKjWkVcRqZR88xfeE2zhaXM3pwJ353dw/tcK+D/uuIw+sTGcrQfhGczC/l0zSNaRVxBt/t\nPMm7n++kqrqGiSOiGXFrJ+1wt4JCXZzCb4d0oX2YP9/tPMnmvaeNLkdE6slssbD42ywWrN6HTzMP\nXhrbh9iY1kaX5TCsCvUZM2YwduxYxowZw5o1a676nHfeeYcJEybU/vz2228zduxYHnroIXbv3g3A\nqVOnmDBhAuPHj2fy5MlUVelPkcQ2PD3ceWpUDM083fnXl/vI05hWEYdTWVXD+0szWbnhGK1a+vDq\nhH5EtmthdFkOpc5Q37RpE1lZWSQnJzN37lymT59+xXOysrJIT0+vbY1s2bKFY8eOkZyczLRp03jr\nrbcAmDVrFhMmTGDhwoW0b9+exYsX2/hwxJW1CfZj/LBIyipqmLNUY1pFHEnxhUpmJm9ny74zdIto\nzquP9KdVkK/RZTmcOkN9wIABzJo1C4DAwEDKysqu+JvgxMREpkyZUvvzhg0biI+PB6BLly4UFxdT\nUlLC5s2bGTJkCABDhgxh/fr1NjsQEYBbb2rDoJjWHMktZsl3h40uR0SskHu2lLc+TCfrRDGx0a14\naWwf/H20w70+6gx1k8mEt7c3AIsWLSIuLu6yzQpLlixh4MCBhIeH1z6Wn59PUFBQ7c9BQUHk5+dT\nXl6Op+fF/0cFBweTl5dnswMRuWT8sEhatfRh9abjZBzWmFYRe7bvWAHTP9pKXmE5I2/tyO9HROPp\noe1e9WX1f7m0tDRSUlJ47bXXah8rKioiJSWFxx9//Lp39Lrav+kOYNJYLo5p7Ym7m4m5K/ZQqDGt\nInbph925vPPZDsora3jinh7cN7izdrg3kFUz6tatW0dSUhLz58/H39+/9vGNGzdSUFDAuHHjqKio\nIDs7m8TERMLCwi67Cj9z5gxhYWH4+vpSWVmJl5cXp0+fJiys7nv2hoYG1OOwxNWFhgbwuxExzF2a\nwb9S9/P/TfwVbm6m2n8TsSWtqRtjsVj4JHU/yWv24+fjyauPDeCmriFGl+UU6gz1kpISZs6cyYIF\nCwgIuHzhDh8+nOHDhwNw4sQJXnnlFaZOncr27duZPXs2Dz74IJmZmbWBPmjQIFJTUxkxYgSpqakM\nHjy4zgLz8s7X89DE1cVGhbI5I5idB/P5cEUG9wzqSGhogNaU2JTW1I2pqjbzweq9bMw8TUhzbyb/\ntjetmzfTf8OfacgviXWG+qpVqygsLGTSpElYLBZMJhOxsbFERkbWbob7pT59+hATE8PYsWNxd3fn\n9ddfB+C5557j5Zdf5rPPPiM8PJzRo0fXu3CRulwa0/rmB1tI+fYw3+3M5WxxOeHBvtwzqCMDo1sZ\nXaKISykpq2L24l0cyCmiS3ggzz3Qi0BfL6PLciomi51/ua3f3qShlnx3mOXrj17x+FMjYxTs0mC6\nUrfO6YILvPv5Tk4XlNE/Kown7+mBl6e70WXZpYZcqWuLoTi97Qev/lcWKzcca+JKRFzTgexC3vpw\nK6cLyrhnUAeeHhWjQG8kVm2UE3FkJ/MvXOPx0iauRMT1bNxzin+u3IvZDI/dFcXtvcPrfpHUm67U\nxemFh1z9rlRmi4WZn25n+4E8zGa7/hZKxOFYLBaWrz9K0rI9eHq4Mfm3vRXoTUBX6uL07hnUkfeX\nZV7xeHiwL3uPFbD3WAEhzb25s28Eg3u30axmkQaqrjHzry/38cPuUwQHNmPSmN60DfWv+4XSYNoo\nJy5h057TrNxwjNyzpbQJ9uOeQR0YGN2KnDMlrN2Ww4aMU1RWm/HydONXMa0Z2i9CJyGxijbKXa60\nvIr/TdnNvuOFdGwdwAsP9KK5fzOjy3IoDdkop1AXl3KtE3BJWRXrdp3k660nOFtcDkCPDi2J7xdB\n764htTeuEfklhfpP8grLeHfRTnLPXqBvZCi/HxFNM22Iu2EKdREr1XUCNpst7DiUT1p6NvuOFwKo\nNS/XpVC/KOtEEX9fvIvzF6oYPqAdY+7oql+G60mhLmKlGzkBqzUv1lCoQ/q+M8xdsYfqGjPjEyIZ\n0jfC6JIcmkJdxEr1OQGXlFXx/a5c1m7NUWteruDKoW6xWPhy03EWfZNFMy93nhnVk15dgo0uy+Ep\n1EWs1JATsNlsYeehfNK25rD3WAGg1ry4bqhX15j5eM0Bvt1xkpYBzXjhgV60b6XBNragUBexkq1O\nwDl5JazdenlrftCPrfkIteZdiiuG+oXyav5vaQaZR87RvpU/LzzQm5YB2uFuKwp1ESvZ+gR8qTX/\n9bYc8ot+as0P7RfBzWrNuwRXC/WzReW8+8VOTuSV0rtLME+NisHbS7c8sSWFuoiVGusErNa863Kl\nUD+SW8zfv9hFUWklQ/tF8NDQbvrFtREo1EWs1BQn4Jy8Er7emsP6S615DzcG9VRr3lm5SqhvP5DH\n+8syqao2Mza+Gwn92xldktNSqItYqSlPwKXlVazbeXlrPqp9C+L7t1Nr3ok4e6hbLBbWpOfw2dqD\neHq68dTIGPp0CzW6LKemUBexkhEn4Gu15of0bcvgXuH4+6g178icOdRrzGY+TTvI19tO0Nzfixce\n6EXH1oFGl+X0FOoiVjL6BFzbms88RWXVxdZ8bExr4vtFEBGm1rwjMnpNNZayimreX5bJrqyzRIT6\nMWlMb4ICvY0uyyUo1EWsZC8nYLXmnYe9rClbOldczqwvdpF9poSenYJ45r6e+DTTDvemolAXsZK9\nnYDNZgs7s/JJS/+pNR8c6M2d/dSadxT2tqYa6vjp87y7aCeFJZXccXM444ZF4u7mZnRZLkWhLmIl\nez4Bn8grYe22E6zPyFVr3oHY85q6UTsP5TNnaSaVVTWMGdKV4QPaYTKpa9TUFOoiVnKEE/C1WvND\n+7WjTze15u2NI6wpa6zdmsMnaQfwdHfj9yOi6dc9zOiSXJZCXcRKjnQCvmZrvm9bBvdWa95eONKa\nuhqz2cLn/znEV1uyCfT15PkHetM5XDvcjaRQF7GSo56A1Zq3X466pgAqKmtIWp7J9oP5hIf4MemB\nXoS08DG6LJenUBexkiOfgOFia/7SGNhftuZv7hasDU0GcNQ1VVhSwawvdnHs1Hl6dGjJ/xvdE1/d\nztguKNRFrOSoJ+BfutSaX7s1hz1HL7Xmm/14r3m15puSI66pnDMlvPvFTs4VV3BbrzY8Mrw7Hu76\nhdBeKNRFrOSIJ+C6nMgv5eutOfxwWWu+FUP7taOdWvONztHWVMaRs/xjSQbllTX8Jq4zd8d20A53\nO6NQF7GSo52Ab8SF8irWXbU1H8HN3ULUmm8kjrSmvt1xgo9SD+DmZuLJe3swoEcro0uSq1Coi1jJ\nkU7A9WU2W9iVdZa0rdlqzTcBR1hTZouFxd9ksXrTcfx9PHn+N73oGtHc6LLkGhTqIlZyhBOwLf2y\nNe/p4cYgteZtyt7XVGVVDXNX7GHr/jxaBfkyeUwvwlr6Gl2WXIdCXcRK9n4CbiwXLu2a35ZDXuHF\n1nz3di2I76/WfEPZ85oqLq3k74t3cfhkMZHtWvDs/TepU+MAFOoiVrLnE3BTuNSaX7s1m8yfteaH\n9I3gdrXm68Ve19SJ/FJmLdpJflE5g2Ja89hdUXh66Jc3R6BQF7GSvZ6AjXCpNb8+4xQVVTV4ergR\nG92Kof0iaN+q/icVV2OPa2rv0XPMXpJBWUU1o27rxMhbO2qHuwNRqItYyR5PwEa7Vmt+aL8I+kSq\nNV8Xe1tT63ad5MMv9wPwu7t7MKhna4MrkhulUBexkr2dgO2J2Wxh1+GzrE3/qTUf9OOuebXmr81e\n1pTFYmHJusOsWH8MP28Pnr3/Jrq3b2l0WVIPCnURK9nLCdjencwvZe22HNbvVmu+Lvawpqqqa/jn\nqn1s2nOasBY+TPptb1oHaYe7o2r0UJ8xYwbbtm2jpqaGiRMnkpCQUPtvn3/+OYsXL8bd3Z2oqChe\nf/11LBYLb7zxBgcOHMDLy4s///nPdOrUiVdeeYWMjAxatrz42+MTTzxBXFzcdT/b6P+xiHOxhxOw\nI1Frvm5Gr6nzFyp5L2U3h3KK6BrRnOfuv4kAXy/D6pGGa0ioe9T1hE2bNpGVlUVycjKFhYWMHj26\nNtTLy8tZvXo1n376KW5ubjz66KPs2LGD/Px8SkpKSE5OJjs7m7feeos5c+YA8NJLL9UZ5CJiH3y9\nPRk2oD3x/dtdbM1vzSHzyDn2ZxcSFNiMIX3acnvvcIWIQU6du8C7n+/kTGEZA3qE8cQ9PfD0cDe6\nLDFQnaE+YMAAevfuDUBgYCBlZWVYLBZMJhPe3t588MEHAJSVlVFSUkJISAjp6en06tULgHbt2nHy\n5EnsvMsvItfh5mbi5q4h3Nw15LLW/OJvD7Psh6NqzRtg//ECZqfsprS8mnt/1YH7BnfGTTvcXV6d\nvbNL4Q2waNEi4uLirvjTiKSkJIYNG8Zdd91FREQEkZGRrFu3DrPZzOHDh8nOzqag4OLGm4ULF/Lo\no4/y4osvUlhY2AiHJCKNKTzEjwnDuvPO//sVY4d2o4W/F+t25fLmB1tI/Hgb6fvOUGM2G12mU9uQ\neYp3PttBeWUNj98dxf23d1GgC3ADG+XS0tKYO3cu8+fPx9//yttLVlZW8uSTTzJ58mT69OnDrFmz\n2LRpE927d2f37t28//77HDx4kBYtWhAVFUVSUhKnT5/mtddeu+7n6vtPsSWjv/90RmaLhd1ZZ0n7\nsTUPuFRrvinXlMViYfkPR/n390fwaebBs6N70qNjUJN8tjSdRt8ot27dOt577z3mz59PQMBPH1ZU\nVMTBgwfp378/APPmzcNkMvHEE09c9vqEhATWrFlz2WNZWVm8+eabfPTRR/UuXkTsS/bp86z84Qhr\ntxynvLIGLw834vpGcO9tnencVgNEGqKquob3Pt/Bf7bmEBbkyxtPDKR960CjyxI7U+d36iUlJcyc\nOZMFCxZcFugA1dXVTJ06leXLl+Pj48OuXbu477772LdvHx9++CHTp0/nu+++IyYmBoDnn3+eP/zh\nD7Rr145NmzYRGRlZZ4G6qhJb0pV64/J2g98M7sRdt7Tj+925fL01hzWbj7Nm83Ei27Ug3gl3zTfF\nmiopq+J/U3azP7uQTm0Cef6BXvi4m7SWnVSj7n5ftWoVhYWFTJo0qXaDXGxsLJGRkcTHx/Pss88y\nYcIEPDw8iIqK4s4778RisWCxWBgzZgze3t785S9/AWDcuHFMnjwZHx8f/Pz8mD59er0LFxH75evt\nwbBb2hHfP+Ky1vwB7Zq/YWcKLvDuol2cOneBft1DefLeaJp5aoe7XJ1uPiMuRVfqxsk9W8rarTn8\n8OMNbTzc3YiNaUW8g++ab8w1dSiniL8v3kVJWRV3DWzPb+7QhjhXoDvKiVhJoW68C+XV/LA7l7Vb\nczhTWAZAZERz4vu3c8jWfGOtqc17TzNvxV7MZgvjh0dyx81tbf4ZYp8U6iJWUqjbj0u75tduzSHj\nx13zLQOacWdfx2rN23pNWSwWVm08xuJvD+Pt5c5/3deTnp2Dbfb+Yv8U6iJWUqjbp9rWfMYpKip/\nbM1HtyK+v/235m25pqprzHyUup91u3IJCmzGpAd6ExF25Z8Qi3NTqItYSaFu367Vmh/avx197bQ1\nb6s1daG8iv9dksHeYwV0aB3ACw/0ooV/MxtUKI5GoS5iJYW6YzBbLGQcPktauv235m2xpvILy3j3\ni12czC+lT7cQJo6IoZmXdri7KoW6iJUU6o4n92wpX289wfcZuZe15of2i6BDa+Nb8w1dU4dPFvP3\nxbsoLq0koX87HryzK25u2uHuyhTqIlZSqDuu2tb8thzOFFxszXe7tGu+Wwge7sa05huyprbuP8Pc\n5XuoqjHzcHwkQ/tF2Lg6cUQKdRErKdQd37Va80P6tOX2m8MJbOLWfH3WlMViIXVzNov+cwgvT3ee\nHhVD764hjVShOBqFuoiVFOrOxR5a8ze6pmrMZj5ec5Bvtp+ghb8XLzzQ2y6+RhD7oVAXsZJC3TmV\nVVTz/aVd803cmr+RNVVWUc3/Lc0g4/A52oX588IDvQgK9G602sQxKdRFrKRQd24XW/PnSNuaTcbh\npmnNW7umzhWX8+6ineTkldKrSzBPjYzBp1md4zfEBSnURaykUHcdV2vND4wOI75fO5u2u61ZU8dO\nnefdL3ZSVFLJkL5teTi+m13+zb3YB4W6iJUU6q7nUmv+6605nP5Za35ovwj6RoY2uDVf15racTCf\nOcsyqKoy8+CdXUm4pR0mDWWR61Coi1hJoe66rtWav6NPW+Ia0Jq/3ppak55N8tqDeLq7MXFkDH0j\nQ+tdv7gOhbqIlRTqAj+25red4PvdDW/NX21Nmc0WPl17kLVbc2ju58XzD/SiU5tAWx6CODGFuoiV\nFOryc2UVP91r/lJrvmtEc+JvoDX/yzVVXlnN+0sz2Zl1lrYhfrwwphchzX0a7RjE+SjURaykUJer\nudSaX7s1h92HzwLWt+Z/vqYKzlcw64udHD9dQkzHljxz3034emuHu9wYhbqIlRTqUpdT5y5cHAO7\nO5fyS635HmHE9796a/7Smjp++jyzvthFwfkKbu8dzvhhkYbdulYcm0JdxEoKdbFWbWt+2wlOn7sA\nXN6a37o/j5UbjnLy7AWCAppRWFJBdY2FMXd04dcD22uHu9SbQl3ESgp1uVFmi4XMI+dIS/+pNe/b\nzJ0LFTVXPDehfwQPxUc2dYniZBoS6vqyR0TkOtxMJm7qHMxNnYM5de4CX2/NYe22nKs+d++xwiau\nTuRy+sJHRMRKrYN8eTghkms11nPPljZpPSK/pFAXEblB4SF+V328TfDVHxdpKgp1EZEbdM+gjtd4\nvEPTFiLyC/pOXUTkBg2MbgXAyg3HyD1bSptgP+4Z1KH2cRGjaPe7uBTtfhdb05oSW2vI7ne130VE\nRJyEQl1ERMRJKNRFRESchEJdRETESSjURUREnIRCXURExEko1EVERJyEQl1ERMRJKNRFRESchFW3\niZ0xYwbbtm2jpqaGiRMnkpCQUPtvn3/+OYsXL8bd3Z2oqChef/11LBYLb7zxBgcOHMDLy4s///nP\ndOrUiVOnTvGHP/wBi8VCaGgoM2bMwNPTs9EOTkRExJXUeaW+adMmsrKySE5OZu7cuUyfPr3238rL\ny1m9ejWffvopn3zyCVlZWezYsYO1a9dSUlJCcnIyb731Fv/zP/8DwKxZs5gwYQILFy6kffv2LF68\nuPGOTERExMXUGeoDBgxg1qxZAAQGBlJWVsal28V7e3vzwQcf4ObmRllZGSUlJYSEhHD06FF69eoF\nQLt27Th58iRms5nNmzczZMgQAIYMGcL69esb67hERERcTp2hbjKZ8Pb2BmDRokXExcVhMpkue05S\nUhLDhg3jrrvuIiIigsjISNatW4fZbObw4cNkZ2dTUFBAWVlZbbs9ODiYvLy8RjgkERER12T16NW0\ntDRSUlKYP3/+Ff82ceJEHnvsMZ588kn69evH7bffzvbt2xk/fjzdu3enS5cu/HIYnLXD4RoyrUbk\narSmxNa0psReWBXq69atIykpifnz5+Pv71/7eFFREQcPHqR///54eXlx++23s23bNvr06cMLL7xQ\n+7xhw4YRHByMn58flZWVeHl5cfr0acLCwmx/RCIiIi6qzvZ7SUkJM2fOZM6cOQQEXP7baHV1NVOn\nTqWsrAyAXbt20alTJ/bt28ef/vQnAL777juio6MxmUwMGjSI1NRUAFJTUxk8eLCtj0dERMRl1Xml\nvmrVKgoLC5k0aRIWiwWTyURsbCyRkZHEx8fz7LPPMmHCBDw8PIiKiuLOO+/EYrFgsVgYM2YM3t7e\n/OUvfwHgueee4+WXX+azzz4jPDyc0aNHN/oBioiIuAqTxdovt0VERMSu6Y5yIiIiTkKhLiIi4iQU\n6iIiIk5CoS4iIuIkDA/1AwcOkJCQwMcff1z72Ntvv83YsWN56KGH2L1792XP37ZtG3/84x+ZMmUK\nmZmZTV2uOIAbXVM7duzg1Vdf5ZVXXmHPnj1NXa7YuRtdT3l5eUyaNIkvvviiqUsVB1LXusrIyAAu\n/qn4q6++yp/+9Cdyc3PrfF9DQ72srIxp06YxaNCg2se2bNnCsWPHSE5OZtq0abz11luXvSYgIIBp\n06bx2GOPsXnz5qYuWexcfdaUr68vb7zxBo8++ijp6elNXbLYsfqsJzc3Nx588MGmLlUciDXratq0\naQAkJyfz5ptv8swzz/D555/X+d6GhnqzZs2YN2/eZXeW27BhA/Hx8QB06dKF4uJiSktLa/+9W7du\nbNiwgb/+9a+1zxO5pD5rKjIyksrKSj755BPuu+++Jq9Z7Fd91lNwcDDu7u5NXqs4jhtZV9XV1Xh6\nehIWFsbZs2frfG9DQ93NzQ0vL6/LHsvPzycoKKj256CgIPLz81m0aBHTpk1j165dxMXF8be//Y0F\nCxY0ccVi7+qzpi7dNfHFF18kMDCwqUsWO1af9XSJbgEi13Ij68rHx4fKykpOnTpFeHh4ne9t9UAX\no5jNZgDGjBkDXLwP/euvv05ZWRkjR440sjRxUL9cU3/7298oLS3lH//4B/379ychIcHI8sTB/HI9\nbdiwgU8//ZTS0lJatmypjqLUy6V1NXbsWN58803MZjOTJ0+u83V2F+phYWHk5+fX/nzmzBlCQ0Nr\nfx48eLDuGS83pK41Zc3/UEQuqWs9DRo06LLvSkWsca115evry/Tp061+H8N3v//SrbfeWjv0JTMz\nk1atWuHr62twVeLItKbElrSepDHYal0ZeqWemZlJYmIiJ0+exMPDg9TUVGbPnk10dDRjx47F3d2d\n119/3cgSxcFoTYktaT1JY2jMdaWBLiIiIk7C7trvIiIiUj8KdRERESehUBcREXESCnUREREnoVAX\nERFxEgp1ERERJ6FQFxERcRIKdRERESehUBcREXES/z8ovYMf4dNNtQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0xaa473c8c>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "Cs = np.logspace(-3,0,4)\n",
    "res = []\n",
    "for C in Cs:\n",
    "    res.append(score(LogisticRegression(C = C)))\n",
    "plt.semilogx(Cs, res,'-o');"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2.3960927513943577"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "score(LogisticRegression(C=0.02))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>F23-</th>\n",
       "      <th>F24-26</th>\n",
       "      <th>F27-28</th>\n",
       "      <th>F29-32</th>\n",
       "      <th>F33-42</th>\n",
       "      <th>F43+</th>\n",
       "      <th>M22-</th>\n",
       "      <th>M23-26</th>\n",
       "      <th>M27-28</th>\n",
       "      <th>M29-31</th>\n",
       "      <th>M32-38</th>\n",
       "      <th>M39+</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>device_id</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1002079943728939269</th>\n",
       "      <td>0.048425</td>\n",
       "      <td>0.054211</td>\n",
       "      <td>0.048074</td>\n",
       "      <td>0.062302</td>\n",
       "      <td>0.062072</td>\n",
       "      <td>0.050612</td>\n",
       "      <td>0.110474</td>\n",
       "      <td>0.154368</td>\n",
       "      <td>0.086057</td>\n",
       "      <td>0.101091</td>\n",
       "      <td>0.125335</td>\n",
       "      <td>0.096980</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>-1547860181818787117</th>\n",
       "      <td>0.049154</td>\n",
       "      <td>0.050907</td>\n",
       "      <td>0.042537</td>\n",
       "      <td>0.049722</td>\n",
       "      <td>0.060525</td>\n",
       "      <td>0.055544</td>\n",
       "      <td>0.100910</td>\n",
       "      <td>0.150901</td>\n",
       "      <td>0.079551</td>\n",
       "      <td>0.104124</td>\n",
       "      <td>0.131345</td>\n",
       "      <td>0.124780</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7374582448058474277</th>\n",
       "      <td>0.050983</td>\n",
       "      <td>0.044571</td>\n",
       "      <td>0.034040</td>\n",
       "      <td>0.057697</td>\n",
       "      <td>0.085255</td>\n",
       "      <td>0.060924</td>\n",
       "      <td>0.080322</td>\n",
       "      <td>0.115715</td>\n",
       "      <td>0.070647</td>\n",
       "      <td>0.106860</td>\n",
       "      <td>0.145422</td>\n",
       "      <td>0.147565</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>-6220210354783429585</th>\n",
       "      <td>0.045871</td>\n",
       "      <td>0.050323</td>\n",
       "      <td>0.030296</td>\n",
       "      <td>0.058017</td>\n",
       "      <td>0.073379</td>\n",
       "      <td>0.056301</td>\n",
       "      <td>0.079781</td>\n",
       "      <td>0.146809</td>\n",
       "      <td>0.078471</td>\n",
       "      <td>0.115130</td>\n",
       "      <td>0.148457</td>\n",
       "      <td>0.117164</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>-5893464122623104785</th>\n",
       "      <td>0.045344</td>\n",
       "      <td>0.061370</td>\n",
       "      <td>0.042611</td>\n",
       "      <td>0.062047</td>\n",
       "      <td>0.057447</td>\n",
       "      <td>0.046417</td>\n",
       "      <td>0.083700</td>\n",
       "      <td>0.157953</td>\n",
       "      <td>0.094920</td>\n",
       "      <td>0.105024</td>\n",
       "      <td>0.142804</td>\n",
       "      <td>0.100364</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                          F23-    F24-26    F27-28    F29-32    F33-42  \\\n",
       "device_id                                                                \n",
       " 1002079943728939269  0.048425  0.054211  0.048074  0.062302  0.062072   \n",
       "-1547860181818787117  0.049154  0.050907  0.042537  0.049722  0.060525   \n",
       " 7374582448058474277  0.050983  0.044571  0.034040  0.057697  0.085255   \n",
       "-6220210354783429585  0.045871  0.050323  0.030296  0.058017  0.073379   \n",
       "-5893464122623104785  0.045344  0.061370  0.042611  0.062047  0.057447   \n",
       "\n",
       "                          F43+      M22-    M23-26    M27-28    M29-31  \\\n",
       "device_id                                                                \n",
       " 1002079943728939269  0.050612  0.110474  0.154368  0.086057  0.101091   \n",
       "-1547860181818787117  0.055544  0.100910  0.150901  0.079551  0.104124   \n",
       " 7374582448058474277  0.060924  0.080322  0.115715  0.070647  0.106860   \n",
       "-6220210354783429585  0.056301  0.079781  0.146809  0.078471  0.115130   \n",
       "-5893464122623104785  0.046417  0.083700  0.157953  0.094920  0.105024   \n",
       "\n",
       "                        M32-38      M39+  \n",
       "device_id                                 \n",
       " 1002079943728939269  0.125335  0.096980  \n",
       "-1547860181818787117  0.131345  0.124780  \n",
       " 7374582448058474277  0.145422  0.147565  \n",
       "-6220210354783429585  0.148457  0.117164  \n",
       "-5893464122623104785  0.142804  0.100364  "
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf = LogisticRegression(C=0.02, multi_class='multinomial',solver='lbfgs')\n",
    "clf.fit(Xtrain, y)\n",
    "pred = pd.DataFrame(clf.predict_proba(Xtest), index = gatest.index, columns=targetencoder.classes_)\n",
    "pred.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def evaluate"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [Root]",
   "language": "python",
   "name": "Python [Root]"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}