alex.ipynb

{
 "metadata": {
  "name": ""
 },
 "nbformat": 3,
 "nbformat_minor": 0,
 "worksheets": [
  {
   "cells": [
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "import csv\n",
      "import pandas as pd\n",
      "import matplotlib as plt\n",
      "import ast\n",
      "import sklearn.linear_model as lm\n",
      "\n",
      "import numpy as np\n",
      "import scipy as sp\n",
      "import scipy.stats as st\n",
      "import pickle\n",
      "from sklearn import cross_validation\n",
      "from sklearn import datasets\n",
      "from sklearn.metrics import confusion_matrix, roc_curve, auc, f1_score\n",
      "from sklearn.externals import joblib\n",
      "from sklearn.naive_bayes import BernoulliNB, MultinomialNB, GaussianNB\n",
      "from sklearn.cross_validation import train_test_split\n",
      "import matplotlib.pyplot as plt\n",
      "from sklearn.feature_extraction.text import TfidfVectorizer\n",
      "\n",
      "import json\n",
      "import nltk\n",
      "import unicodedata\n",
      "\n",
      "from sklearn.feature_extraction.text import CountVectorizer"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 1
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "df = pd.DataFrame.from_csv('train.tsv', sep='\\t')\n",
      "for idx, i in enumerate(df.boilerplate):\n",
      "    try:\n",
      "        df.boilerplate.iloc[idx] = ast.literal_eval(i)\n",
      "    except:\n",
      "        temp = df.boilerplate.iloc[idx].split('null,')\n",
      "        fixed = ''\n",
      "        if len(temp) == 2:\n",
      "            fixed = (temp[0]+'\"empty\",'+temp[1])\n",
      "        else:\n",
      "            fixed = (temp[0]+'\"empty\",'+temp[1] + '\"empty\",' + temp[2])\n",
      "        df.boilerplate.iloc[idx] = ast.literal_eval(fixed)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 2
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# pd.value_counts(df['alchemy_category']).plot(kind='bar');"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 3
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# pd.value_counts(df['alchemy_category'][df['label']==1]).plot(kind='bar');"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 4
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# ratios = pd.value_counts(df['alchemy_category'][df['label']==1])/pd.value_counts(df['alchemy_category'])\n",
      "# print ratios"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 5
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "labels = np.array(df['label'])\n",
      "ltest_list = unique(labels)\n",
      "\n",
      "# print sum(df['label'])"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 6
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# print df[df.columns[2:-1]]\n",
      "# print df.columns"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 7
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# vectorizer = CountVectorizer(min_df=1)\n",
      "vectorizer = TfidfVectorizer(min_df=3,strip_accents=None,ngram_range=(1,2),\n",
      "                             stop_words='english',max_features=None,binary=False,\n",
      "                             norm='l2',use_idf=True,smooth_idf=True,sublinear_tf=True)\n",
      "\n",
      "features = []\n",
      "for i in df.boilerplate:\n",
      "    features.append(i['body'])\n",
      "\n",
      "features = np.array(features)\n",
      "\n",
      "labels = np.array(df['label'])\n",
      "ltest_list = unique(labels)\n",
      "X = vectorizer.fit_transform(features,y=labels)\n",
      "\n",
      "x_train, x_split, labels_train, labels_split = train_test_split(X,labels,test_size=0.4)\n",
      "x_cross, x_test, labels_cross, labels_test = train_test_split(x_split,labels_split,test_size=0.5)\n",
      "\n",
      "\n",
      "minwrong = 1000\n",
      "bestalpha = 0\n",
      "for atest in np.linspace(0,2,101):\n",
      "    mnb = MultinomialNB(alpha=atest)\n",
      "    lpred_cross = mnb.fit(x_train,labels_train).predict(x_cross)\n",
      "    numwrong = (labels_cross != lpred_cross).sum()\n",
      "    if numwrong < minwrong:\n",
      "        minwrong = numwrong\n",
      "        bestalpha = atest\n",
      "print(\"The best alpha is %f\" % (bestalpha))\n",
      "print(\"Number of mislabeled points: %d\" % minwrong)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "The best alpha is 1.740000\n",
        "Number of mislabeled points: 327\n"
       ]
      }
     ],
     "prompt_number": 9
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "\n",
      "mnb = MultinomialNB(alpha=bestalpha)\n",
      "lpred_test = mnb.fit(x_train,labels_train).predict(x_test)\n",
      "print(\"alpha = %f\" % (bestalpha))\n",
      "print(\"Number of mislabeled points : %d\" % (lpred_test != labels_test).sum())\n",
      "\n",
      "cm_mnb = confusion_matrix(labels_test,lpred_test,labels=ltest_list)\n",
      "fig, ax = plt.subplots(figsize=(10,10))\n",
      "clb = ax.matshow(cm_mnb,cmap='Blues')\n",
      "fig.colorbar(clb)\n",
      "ax.set_xlabel('Predicted label')\n",
      "ax.set_ylabel('True label')\n",
      "ax.set_xticks(range(2))\n",
      "ax.set_yticks(range(2))\n",
      "ax.set_xticklabels(ltest_list)\n",
      "ax.set_yticklabels(ltest_list)\n",
      "ax.set_title('Confusion Matrix');\n",
      "print cm_mnb"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "alpha = 1.740000\n",
        "Number of mislabeled points : 328\n",
        "[[572 170]\n",
        " [158 579]]"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n"
       ]
      },
      {
       "metadata": {},
       "output_type": "display_data",
       "png": "iVBORw0KGgoAAAANSUhEUgAAAjIAAAI8CAYAAADrxkEpAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRU5ErkJggg==\n",
       "text": [
        "<matplotlib.figure.Figure at 0x113cadf90>"
       ]
      }
     ],
     "prompt_number": 10
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "probs_mnb = mnb.fit(x_train,labels_train).predict_proba(x_test)\n",
      "fpr_mnb = {}\n",
      "tpr_mnb = {}\n",
      "auc_mnb = {}\n",
      "\n",
      "for i in range(len(ltest_list)):\n",
      "    fpr_sk, tpr_sk, thresholds = roc_curve((labels_test==ltest_list[i]), probs_mnb[:,i])\n",
      "    fpr_mnb[ltest_list[i]] = fpr_sk\n",
      "    tpr_mnb[ltest_list[i]] = tpr_sk\n",
      "    try:\n",
      "        auc_mnb[ltest_list[i]] = auc(fpr_sk,tpr_sk)\n",
      "    except:\n",
      "        auc_mnb[ltest_list[i]] = 0\n",
      "    \n",
      "    plt.figure(1,figsize=(10,8))\n",
      "    plt.plot(fpr_mnb[ltest_list[i]],tpr_mnb[ltest_list[i]],label=ltest_list[i])\n",
      "    plt.legend(loc='best')\n",
      "for item in auc_mnb.iteritems():\n",
      "    print item"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "(0, 0.85449132675266148)\n",
        "(1, 0.85449132675266148)\n"
       ]
      },
      {
       "metadata": {},
       "output_type": "display_data",
       "png": "iVBORw0KGgoAAAANSUhEUgAAAlYAAAHfCAYAAACS8paSAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
       "text": [
        "<matplotlib.figure.Figure at 0x113cadf50>"
       ]
      }
     ],
     "prompt_number": 13
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "x_train, x_split, labels_train, labels_split = train_test_split(X,labels,test_size=0.4)\n",
      "x_cross, x_test, labels_cross, labels_test = train_test_split(x_split,labels_split,test_size=0.5)\n",
      "\n",
      "rd = lm.LogisticRegression(penalty='l1')\n",
      "rd.fit(x_train,labels_train)\n",
      "lpred_test = rd.predict(x_test)\n",
      "\n",
      "cm_mnb = confusion_matrix(labels_test,lpred_test,labels=ltest_list)\n",
      "fig, ax = plt.subplots(figsize=(10,10))\n",
      "clb = ax.matshow(cm_mnb,cmap='Blues')\n",
      "fig.colorbar(clb)\n",
      "ax.set_xlabel('Predicted label')\n",
      "ax.set_ylabel('True label')\n",
      "ax.set_xticks(range(2))\n",
      "ax.set_yticks(range(2))\n",
      "ax.set_xticklabels(ltest_list)\n",
      "ax.set_yticklabels(ltest_list)\n",
      "ax.set_title('Confusion Matrix');\n",
      "print cm_mnb\n",
      "\n",
      "probs_mnb = mnb.fit(x_train,labels_train).predict_proba(x_test)\n",
      "fpr_mnb = {}\n",
      "tpr_mnb = {}\n",
      "auc_mnb = {}\n"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "[[652  75]\n",
        " [230 522]]\n"
       ]
      },
      {
       "metadata": {},
       "output_type": "display_data",
       "png": "iVBORw0KGgoAAAANSUhEUgAAAjIAAAI8CAYAAADrxkEpAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzt3Xv4XVV95/H3J0iCCl4AS4XEgmPkkVLlVqiCgtMbWgfU\nGSvWUpCOM6NWHKfVVttabXXGTls19dL20XJVo6NVCor3C0UuRhMgYLCdOAUJcvFerGKDfOePsxMP\nP/I7v19CzmWvvF885+HsvdfZe+34CF8+a+21U1VIkiT10ZJpd0CSJGlHWchIkqTespCRJEm9ZSEj\nSZJ6y0JGkiT1loWMJEnqrftNuwOSJGk8kkx8jZWqyiSvZyEjSVLD9jjsRRO71p1Xv3Vi19rCQkaS\npJal7Vkkbd+dJElqmoWMJEnqLYeWJElqWSY693biTGQkSVJvmchIktQyJ/tKkiTNJhMZSZJa5hwZ\nSZKk2WQiI0lSy5wjI0mSNJtMZCRJaplzZCRJkmaThYwkSeoth5YkSWqZk30lSZJmk4mMJEktc7Kv\npElJcv8kb0nytSTvvQ/neW6Sj+3Mvk1DkouTnDrtfkiaXRYy0g5I8pQklyS5Pcmt3b9wj90Jp/5P\nwPHAyqp69o6epKreVVW/vBP6cw9JTkhyd5IPzNn/uG7/ZxZ5nlcnOX+hdlX11KpasJ2kEbJkcp8p\nsJCRtlOS5wPnAG8EfgpYCZwF7HDhMeSngPVV9a874Vzj8nXg55LsPbTvNOCfgNoZF0hnZ5xLUtss\nZKTtkGQv4E+BF1bVBVX1g6q6o6reX1Vndm2WJXlTkpu7zxuTLO2OnZBkU5L/kuT/dUNIp3fHXgP8\nIfCsJHckOWNucpHkwC75WNJtn5zkM0m+053v17r9pye5dOh3T0jyha7dmiSPHzr22SS/l+RjSb7d\n/X2fEX8M/wZcAJzS/X434FeBdwFbi48kq5J8Ncm/JPlikuO6/ScCrwCe3d3nVUP9eGWSjwPfAR7Z\n7fvN7vhfJXn/0Pn/NMknt+N/PmnXlEzuMwUWMtL2ORTYC/jgiDa/DxwLPK77PB74g6HjPwEcBhwD\nvAp4a5IHV9UfAf8TeE9V7VVVZzEi4UiyO7AK+L2qekh3nau30W5v4MPAm4G9u998OMlDh5q9EPjf\n3f09FPidEfcHcD7wG933XwauA742p82a7v4fAXwaeF+SpVX10Tn3efjQb17U9W8f4Mbu/rf8GfwP\n4GeSnJbkicAZQ32QtIuykJG2zwrgG1V194g2zwXeXlXfqKpvAO8AhiesLgFeVVVfZzBEdTdwcHcs\nDKUac77PVcBS4FFJHlBVt1XVhm20+xXg+1V1XlXdXVXvAr4P/Ieh81xQVZ+qqpuBDzAotOa/cNUV\nwN5JHs2gmDh3G23eVVXfrqrvAH8CPHjEfW7px0er6sNVdVdV3TXnfD9g8Of4RgaF1G9V1dziSdJc\nzpGRNOQmYN8tQzvzeDiwdmh7LbD/0PYtXYFD9y/rbwB7bm9Hut/+RwYThG9O8qEkK7fRdH/gqjn7\nvggcMLQ9nOTcusj+nA+8GDiBQUJ1j8Ikye8k2ZDkO8BXgWXAvguc8/OjDlbVGuD/dZvvW0QfJTXO\nQkbaPtcB/wI8Y0SbrwFHDW0fBdy8g9f7HrDf0PbwMAxVdUVVPYNB8XQjgyGbbfXniDn77kuftngn\n8ALgw1V15/CBbujnNcCvd8NejwB+wI+LneLeiQzAXdvYN3zeFzFIob4GvPw+9V7aVThHRtIWVXUH\n8LvAW7qJtg9I8qAkz0iyqmu2Gjgjyb5J9mUwl+OdO3jJqxk8IXREkoMZzCEBIMlPdH14ILAbsJlt\nJx4XA/dPcmqS+3UTgh8AfGiozXb/E6iq/hl4EoM5QXN9n8Gk4O8m2Y9BgbVs6PitwCFJls353bb6\nEYBuGOtPGAzd/Qbw8iSP295+S2qLhYy0narqHcDzGEw+vZHBY8dnMChgAF4LXAms7z6f7/ZtPcWo\n0w8fr6pPAucBn+rO/46h40uAlzJIVr7MYCLvf5t7nqr6JvA04CUMhrFeCjytqr41T5+Khfu4pX+X\nV9Wt27jmWuBtDCb5/gODJOumoXO8D/hX4PYkX5ynH1v3dU9GnQ+8vqquraqNwCuB87tJz5J2Uana\nKcs+SJKkGZOk9njiH03sende+hqqaqJjTCYykiSpt3xppCRJLZvSY9GT0vbdSZKkppnISJLUsiVt\nv7bMREaSJPWWiYwkSS1zjowkSdJsMpGRJKllU3p1wKSYyEiSpN6ykJEkSb3l0JIkSS1zsq8kSdJs\nspDRdkvypCTrkqxP8uJp90faVSQ5K8ltSa6ddl/UI8nkPlNgIaPtkmQ34CzgmcCRwG8mecx0eyXt\nMs4GTpx2J6RZ4hwZba+jgY1VdQNAkvcAJwPXT7NT0q6gqi5NcuC0+6GecY6MdA8HADcNbW/q9kmS\ndC9JHpjk3CRXJdmQ5JgkeyW5oJui8MEkew61P7Pbvy7JcQud30JG26um3QFJ0naY/hyZtwGXVNXh\nwGOBLwN/CFxeVY8FrgT+YNDVHAKcwWDqwjOBc5LRkZKFjLbXzcCKoe0VDFIZSZLuIcmDgSdW1VkA\nVXVXVX0XOAk4t2t2LvD07vvJwOqq2txNYdjIYErDvCxktL2+CKxMcmCSpcCzgQun3CdJ0nyyZHKf\nezsI+HqSc5Jcl+TtSR4A7FdVt3VtbgP2677vzz3/43jB6QsWMtouVXUXg9jvg8Ba4KyqcqKvNAFJ\nVgOXA49OclOS5027T9IC7gf8LPB33d+XAc8ablBVxehpCyOnNPjUkrZbVV0CHD7tfki7mqp6zrT7\noB4a4/ouP/rmRu7+1sZRTTYB36yqiwZdyWrgN4Bbk/xkVd2a5OHA7V37udMXlnf75mUiI0mSdshu\n+zyK3VeeuPUzV1XdCmzsnlRaAvwK8CngIuC0rtlpwAXd9wuBU5IsTXIQsBJYM6oPJjKSJGmcTgPO\nA/YFrgV+l0GQcn6S9cBXgFMBqmpDkrMZTF24Czi9G3qaVxY4LkmSeipJ7fGUN03send+5L9TVRN9\nV4FDS5IkqbccWpIkqWVTepnjpEy1kEniuJYkaZcz6eGXlk09kdnjsBdNuwvaQZtvWcPuDx+54KJm\n2Le/8JZpd0E76LV//Gr+4FWvnnY3tIPuv/uEaxhfGilJkjSbpp7ISJKkMTKRkbZtyZ4jX38haUye\ndPwJ0+6CNDNMZLTDdtvLQkaaBgsZbZfGn1oykZEkSb1lISNJknrLoSVJklrmZF9JkqTZZCIjSVLL\nnOwrSZI0m0xkJElqmXNkJEmSZpOJjCRJLXOOjCRJ0mwykZEkqWExkZEkSZpNJjKSJDXMREaSJGlG\nWchIkqTecmhJkqSWtT2yZCIjSZL6y0RGkqSGOdlXkiRpRpnISJLUMBMZSZKkGWUiI0lSw0xkJEmS\nZpSJjCRJDTORkSRJmlEmMpIktaztQMZERpIk9ZeFjCRJ6i2HliRJapiTfSVJkmaUiYwkSQ0zkZEk\nSZpRJjKSJDXMREaSJGlGmchIktQwExlJkqQZZSIjSVLL2g5kTGQkSVJ/WchIkqTecmhJkqSGOdlX\nkiRpRpnISJLUMBMZSZKkHZTkhiTrk1yVZM2cY7+d5O4kew/tO7Nrvy7JcQud30RGkqSGzUAiU8AJ\nVfWt4Z1JVgC/CNw4tO8Q4AzgSOAA4JNJHl1Vd893chMZSZI0btuqpt4AvHzOvpOB1VW1uapuADYC\nR486sYWMJEktywQ/21bAp7uhpecDJDkZ2FRV6+e03R/YNLS9iUEyMy+HliRJ0g75t1u+xOZbNizU\n7NiquiXJY4CLk3wZeAXwS0NtRo1/1aiTW8hIktSwcc6RWbb/oSzb/9Ct29+/+v33alNVt3R/vz7J\nB4HjgYOAa7q+LQfWJjkGuBlYMfTz5d2+eTm0JEmSxiLJA5Ls1X1/GPBUYE1V7VdVB1XVQQyGj46o\nqtuAC4FTkixNchCwElgz3/nBREaSpKZN+aml/YAPdn34JvDGqvr4nDZbh46qakOSs4G1wF3A6VXl\n0JIkSZq8qvpn4LAF2jxyzvYqYNVir+HQkiRJ6i0TGUmSGjYDC+KNlYmMJEnqLRMZSZIaZiIjSZI0\no0xkJElqWduBjImMJEnqLxMZSZIa5hwZSZKkGWUiI0lSw0xkJEmSZpSFjCRJ6i2HliRJaphDS5Ik\nSTPKREaSpJa1HciYyEiSpP4ykZEkqWHOkZEkSZpRJjKSJDXMREaSJGlGmchIktQwExlJkqQZZSIj\nSVLDTGQkSZJmlIWMJEnqLYeWJElqWdsjSyYykiSpv0xkJElqmJN9JUmSZpSJjCRJDTORkSRJmlEm\nMpIkNazxQMZERpIk9ZeJjCRJDXOOjCRJ0oyykJEkSb3l0JIkSQ1rfGTJREaSJPWXiYwkSQ1zsq8k\nSdKMMpGRJKlhjQcyJjKSJKm/TGQkSWrYkiVtRzImMpIkqbdMZCRJaphzZCRJkmaUiYwkSQ1zHRlJ\nkqQZZSEjSZJ6y0JGkqSGJZP7zN+H7JbkqiQXdduHJPlQkquTXJTkMUNtz0yyPsm6JMctdH9jLWSS\nPKnryPokLx7ntSRJ0sx6CbABqG77VcB5VXUY8O5umySHAGcARwLPBM5JMrJWGVshk2Q34KyuI0cC\nvzlccUmSpPFLMrHPPNdfDjwVeAewpdF3gX26ImUf4Nvd/pOB1VW1uapuADYCR4+6v3E+tXQ0sLHr\nCEne03Xw+jFeU5IkzZY3Ai8DHjS072XAGuD1wC3AUd3+/YErh9ptAg4YdfJxFjIHADfN6cwxY7ye\nJEmaY5yPX3/vhqv53g3XjLr204Dbq+qqJCcMHToLeDPwN8CLuu1fnec0Nc9+YLyFzMgLb7H5ljVb\nvy/Z8wB222tk4SVJUq/8wyWf5R8u+ey0uzEWex54GHseeNjW7dv/4fy5TZ4AnJTkqcAewIOSnA8c\nB5xaVXcl+VvgFV37m4EVQ79f3u2b1zgLmbmdWcEglbmH3R8+cuhLkqRee9LxJ/Ck40/Yuv26P3nN\nRK8/zfXwquqVwCsH/cjxwO9U1alJVgMnAe9lMO3kE91PLgTeneQNDEZ2VjIYgprXOJ9a+iKwMsmB\nSZYCz+46KEmSdk1bRmteCzw9yTUMJgK/DqCqNgBnA2uBDwCnV9V0hpa6uOgM4IPddd5eVU70lSRp\ngmblFQVVdQlwSff9S8Bz5mm3Cli12POO9V1LXacPH+c1JEnSrsuXRkqS1LAZCWTGxlcUSJKk3rKQ\nkSRJveXQkiRJDZuVyb7jYiIjSZJ6y0RGkqSGNR7ImMhIkqT+MpGRJKlhzpGRJEmaUSYykiQ1rPFA\nxkRGkiT1l4mMJEkNc46MJEnSjLKQkSRJveXQkiRJDWt8ZMlERpIk9ZeJjCRJDXOyryRJ0owykZEk\nqWGNBzImMpIkqb9MZCRJaphzZCRJkmaUiYwkSQ1rPJAxkZEkSf1lIiNJUsOcIyNJkjSjLGQkSVJv\nObQkSVLDHFqSJEmaUSYykiQ1rPFAxkRGkiT1l4mMJEkNc46MJEnSjDKRkSSpYY0HMiYykiSpv0xk\nJElqmHNkJEmSZpSFjCRJ6i2HliRJaljjI0smMpIkqb9MZCRJatiSxiMZExlJktRbJjKSJDWs8UDG\nREaSJPWXiYwkSQ1zQTxJkqQZZSIjSVLDlrQdyJjISJKk/rKQkSSpYUkm9hnRh92SXJXkom77z5Jc\nn2RdkjclefBQ2zOTrO+OHbfQ/VnISJKkcXsJsAGobvvjwE8DRwEPBF4BkOQQ4AzgSOCZwDlJRtYq\nFjKSJGlskiwHngq8AwhAVX2iqu6uqruBjwHLu+YnA6uranNV3QBsBI4edX4LGUmSGpZM7jOPNwIv\nA+6e5/jzgb/vvu8PbBo6tgk4YNT9+dSSJEnaId/8x7V885/WzXs8ydOA26vqqiQnbOP47wN3VNX7\nRlymRhyzkJEkqWVhfM9f73vwUex78FFbtzd++B1zmzwBOCnJU4E9gAclOa+qfiPJ6QyGnH5+qP3N\nwIqh7eXdvnk5tCRJksaiql5ZVSuq6iDgFODTXRFzIoPhppOq6s6hn1wInJJkaZKDgJXAmlHXMJGR\nJKlhM7QgXvjxMNGbgaXAJ7vHtq+oqhdW1YYkZwNrgbuA06vKoSVJkjRdVfVZ4LPd95Uj2q0CVi32\nvBYykiQ1zJdGSpIkzSgTGUmSGtZ4IGMiI0mS+stCRpIk9ZZDS5IkNWxJ42NLJjKSJKm3TGQkSWpY\n44GMiYwkSeovExlJkhrmgniSJEkzykRGkqSGNR7ImMhIkqT+MpGRJKlhriMjSZI0o0xkJElqWNt5\njImMJEnqMQsZSZLUWw4tSZLUMBfEkyRJmlEmMpIkNWxJ24GMiYwkSeovExlJkhrmHBlJkqQZZSIj\nSVLDGg9kTGQkSVJ/mchIktSw1ufIzFvIJHnziN9VVZ05hv5IkiQt2qhEZi1Q3fct5Vx132ubv5Ak\nSZqgeQuZqjpneDvJ/avqB2PvkSRJ2ml2+QXxkhyd5MPAhm77sCRvG3vPJEmSFrCYyb5vB54LnA9Q\nVVcnOX6svZIkSTtF65N9F/P49e5Vdd2WjSTLgPuPr0uSJEmLs5hE5p+SnAyQ5BHAi4EvjLVXkiRp\np2g7j1lcIvN84NnAcmAdsD/wwnF2SpIkaTEWTGSq6uvAr02gL5IkaSdbsqvPkUnyyCRvTvKP3ecv\nkxw0ic5JkiSNspihpf8DfBV4InAccCPwvnF2SpIk7RzJ5D7TsJhCZk/gz6vq9m6Y6U3dPkmSpKka\n9a6lvRlMdl4N/EWS87vt5wLvnkz3JEnSfdH6OjKjJvuu457vVHpG9/ct71r643F1SpIkaTFGvWvp\nwAn2Q5IkabstZkE8khwKHALssWVfVZ03rk5JkqSdo/GRpYULmSQvYLAo3nLgMuDngYsACxlJkjRV\ni3lq6VTg54CvV9UzgKOAh421V5IkaadYkkzsM5X7W0Sb3avq34AbkhwAfAVYMd5uSZIkLWwxc2S+\nmOShwLnApcBm4O/G2itJkrRT7PJzZKrqBd3X/5PkI8BDq+qr4+2WJEnSwkYtiHck91xHZvjYvlW1\nbmy9kiRJO8WuvCDeXzBPIdN58k7uiyRJakiSPYBLgGXAncB7q+qN3bHnAS9isLTLh6vqd7v9ZwL/\nGbgLOLOqPjfqGqMWxDthJ9zDgj76HhcIlqbhUS+5YNpdkDQBi3mqZ1yq6s4kT66q7ydZBqxN8iEG\nS7r8GvD4qtqc5GEASQ4BzgCOBA4APpnk0VV193zXmOb9SZKkxlXV97uvezIIUH4IvAD4X1W1uWvz\n9a7NycDqqtpcVTcAG4GjR53fQkaSpIYlmdhnnusvSXINcBvwlu6BoZXAk5KsS3JJkiO65vsDm4Z+\nvolBMjOvRb2iQJIkaa5N163h5uvWjGzTDQs9LsmBwMVJLmNQfzwSOBb4BeDPgX8/3ylGnX+x71o6\nBnhyVb0+ySOAn6yq0T2XJElNW37o0Sw/9McjP2ve+9Z521bVDUkuBo5nkLS8p6p+AFyU5G+S3B+4\nmXsuuru82zevBYeWkvwV8BLgtG7X94C3LfQ7SZI0fUsyuc9cSfZN8pDu+z7AU4BrgQuAp2bgGOAr\nXVFzIXBKkqVJDmIwBDUyOFlMIvMLwMHAWoCq+lZXNUmSJI3ycODcJLsBtwJvqKpPJfks8Fbgywze\nGPDfAKpqQ5KzGdQcdwGnV9V9Hlr6IbD7lo0kjwF22/57kSRJk7atpGRSqupa4Iht7P8RXfGyjWOr\ngFWLvcZiCpnXA58AfqKrkn4JeOliLyBJkjQui3nX0juTrAV+vtv1v6vq+vF2S5Ik7Qy78isKAOie\nUvpX4KJuVyV5hC+OlCRJ07aYoaWL+fEz3A9lsFjNBuDQcXVKkiTtHNOcIzMJixlaukfBkuQpwElj\n65EkSdIi7cgrCj7KYDEbSZI045LJfaZhMXNkfntocxlwHHDp2HokSZK0SIuZI7Pn0Pc7GTyOfcV4\nuiNJkrR4IwuZbiW+varqdybUH0mStBMtafzx63nnyCS5X7fy3hOT7DXBPkmSJC3KqERmDYNlhS9j\n8GbK9wO3dMeqqj4w7s5JkqT7Zkee6umTUYXMlixqb+CfgSPnHLeQkSRJUzWqkHlYkv/B4HXbkiSp\nhxqfIjOykNkNcG6MJEmaWaMKmVur6jUT64kkSdrpdtmnliRJkmbdqETmFybWC0mSNBaNBzLzJzJV\n9c1JdkSSJGl7LeYVBZIkqaeW7KqJjCRJ0qyzkJEkSb3l0JIkSQ3z8WtJkqQZZSIjSVLDGg9kTGQk\nSVJ/mchIktQwH7+WJEmaUSYykiQ1LLQdyZjISJKk3jKRkSSpYc6RkSRJmlEWMpIkqbccWpIkqWEO\nLUmSJM0oExlJkhqWxt9RYCIjSZJ6y0RGkqSGOUdGkiRpRpnISJLUsManyJjISJKk/jKRkSSpYUsa\nj2RMZCRJUm+ZyEiS1DCfWpIkSZpRFjKSJKm3HFqSJKlhjc/1NZGRJEn9ZSIjSVLDltB2JGMiI0mS\nestCRpKkhiWT+9z72tkjyeeTXJ3kyiQv7fbvleSCJOuTfDDJnkO/ObPbvy7JcQvdn4WMJEkai6q6\nE3hyVR0GHA/8ZpKVwB8Cl1fVY4ErgT8ASHIIcAZwJPBM4JwkI2sVCxlJkhq2JJP7bEtVfb/7uiew\nG/BD4CTg3G7/ucDTu+8nA6uranNV3QBsBI4eeX/35Q9HkiRplCRLklwD3Aa8taq+CuxXVbd1TW4D\n9uu+7w9sGvr5JuCAUef3qSVJkho2zpdG/uO6K/indVeObFNVdwOPS3IgcHGSy+YcryQ16hSjzm8h\nI0mSdsjBRzyeg494/NbtD/3tqnnbVtUNSS5mMFfmtiQ/WVW3Jnk4cHvX7GZgxdDPlnf75uXQkiRJ\nGosk+yZ5SPd9H+ApwLXAhcBpXbPTgAu67xcCpyRZmuQgYCWwZtQ1TGQkSWrYlF9R8HDg3CS7AbcC\nb6iqTyVZA5yfZD3wFeBUgKrakORsYC1wF3B6VTm0JEmSJq+qrgWO2Mb+O/jxk0pzj60C5h+jmsNC\nRpKkho1zsu8scI6MJEnqLRMZSZIa1nggYyIjSZL6y0RGkqSGtZ5YtH5/kiSpYSYykiQ1LI1PkjGR\nkSRJvWUiI0lSw9rOY0xkJElSj1nISJKk3nJoSZKkhvmKAkmSpBllIiNJUsPazmNMZCRJUo+ZyEiS\n1LDGp8iYyEiSpP4ykZEkqWG+okCSJGlGmchIktSw1hOL1u9PkiQ1zERGkqSGOUdGkiRpRlnISJKk\n3nJoSZKkhrU9sGQiI0mSesxERpKkhjnZV5IkaUaZyEiS1LDWE4vW70+SJDXMREaSpIY5R0aSJGlG\nmchIktSwtvMYExlJktRjFjKSJKm3HFqSJKlhjc/1NZGRJEn9ZSIjSVLDljQ+3ddERpIk9ZaJjCRJ\nDXOOjCRJ0owaayGT5KwktyW5dpzXkSRJ25YJ/jUN405kzgZOHPM1JEnSLmqsc2Sq6tIkB47zGpIk\naX7OkZEkSZpRU39q6ew3/+nW74cdfSyHH3PcFHsjSdLO9cObr+OHN183teu3vo7M1AuZ5734d6fd\nBUmSxmbZAYey7IBDt27f8cX3TrE37XFoSZIk9da4H79eDVwOPDrJTUmeN87rSZKke0om95mGsRYy\nVfWcqtq/qpZV1YqqOnuc15MkSbMjyYokn0nypSSfTXL6nOO/neTuJHsP7Tszyfok65IsOHF26nNk\nJEnS+Ez58evNwEur6uok+wLXJfl8VV2fZAXwi8CNWxonOQQ4AzgSOAD4ZJJHV9Xd813AOTKSJGks\nqurWqrq6+/4N4AvA/t3hNwAvn/OTk4HVVbW5qm4ANgJHj7qGiYwkSQ2b1qsD5kryKOCngSuTnAxs\nqqr1uWdktD9w5dD2JgbJzLwsZCRJ0g65Zs1lXLPmsgXbJdkTeA/wUqCAVzIYVtraZMTPa9S5LWQk\nSWrYkjEGMocfcyyHH3Ps1u13vu3P7tUmye7A3wHvrKq/T/IzwIHANV0asxxYm+QY4GZgxdDPl3f7\n5uUcGUmSNBYZVCp/C3ypqt4EUFXXVtV+VXVQVR3EYPjoiKq6DbgQOCXJ0iQHASuBNaOuYSIjSVLD\npjxH5ljg14H1Sa7q9r2yqj4y1Gbr0FFVbUhyNrAWuAs4vaocWpIkSZNXVZ9jgdGfqnrknO1VwKrF\nXsOhJUmS1FsmMpIkNWzKC+KNnYmMJEnqLRMZSZIaNisL4o2LiYwkSeotExlJkho2zgXxZoGJjCRJ\n6i0TGUmSGuYcGUmSpBllIiNJUsNcR0aSJGlGmchIktSwxgMZExlJktRfFjKSJKm3HFqSJKlhSxqf\n7WsiI0mSestERpKkhrWdx5jISJKkHjORkSSpZY1HMiYykiSpt0xkJElqmC+NlCRJmlEmMpIkNazx\nZWRMZCRJUn9ZyEiSpN5yaEmSpIY1PrJkIiNJkvrLREaSpJY1HsmYyEiSpN4ykZEkqWEuiCdJkjSj\nTGQkSWqYC+JJkiTNKBMZSZIa1nggYyIjSZL6y0RGkqSWNR7JmMhIkqTespCRJEm95dCSJEkNc0E8\nSZKkGWUiI0lSw1wQT5IkaUaZyEiS1LDGAxkTGUmS1F8mMpIktazxSMZERpIk9ZaJjCRJDXMdGUmS\npBllIiNJUsNcR0aSJGkHJDkryW1Jrh3ad0iSDyW5OslFSR4zdOzMJOuTrEty3GKuYSEjSZLG5Wzg\nxDn7XgWcV1WHAe/utklyCHAGcCTwTOCcJAvWKRYykiQ1LBP8zFVVlwLfnrP7u8A+XZGyz9Dxk4HV\nVbW5qm4ANgJHL3R/zpGRJEmT9DJgDfB64BbgqG7//sCVQ+02AQcsdDILGUmSWjbGyb5fuPxSvnDF\npdv7s7OANwN/A7yo2/7VedrWQiezkJEkSTvkZ5/wRH72CU/cuv1Xb/xfi/nZccCpVXVXkr8FXtHt\nvxlYMdRuebdvJOfISJLUsEzwr0X6DHBS9/1k4BPd9wuBU5IsTXIQsJLBENRIFjKSJGkskqwGLgcO\nTnJTkucvy4eeAAAHY0lEQVQBrwWenuQa4KnA6wCqagODp5zWAh8ATq8qh5YkSdqVTXNBvKp6zjyH\ntrm/qlYBq7bnGiYykiSpt0xkJElqWONvKDCRkSRJ/WUhI0mSesuhJUmSWtb42JKJjCRJ6i0TGUmS\nGrYdC9X1komMJEnqLRMZSZIaNs0F8SbBREaSJPWWiYwkSQ1rPJAxkZEkSf1lIiNJUssaj2RMZCRJ\nUm+ZyEiS1DDXkZEkSZpRFjKSJKm3HFqSJKlhLognSZI0o0xkJElqWOOBjImMJEnqLxMZSZJa1ngk\nYyIjSZJ6y0RGkqSGuSCeJEnSjDKRkSSpYa4jI0mSNKMsZCRJUm85tCRJUsMaH1kykZEkSf1lIiNJ\nUssaj2RMZCRJUm+ZyEiS1DAXxJMkSZpRJjKSJDXMBfEkSZJmlImMJEkNazyQMZHRjrvq85+bdhek\nXdIPb75u2l2QZoaFjHbY1Wsum3YXpF2ShYy2RzK5zzRYyEiSpN6ykJEkSb2VqprexZPpXVySpCmp\nqokMxCSpm771w0lcCoAVey+b2L1tMdWnliZ9s5IkqS0+fi1JUsNcEE+SJGlGmchIktSwxgMZExlp\nUpL8KMlVSdYm+bMku9+Hc52T5D9239+e5DEj2h6f5PE7cI0bkuy92P1z2nxvO6/16iS/vb19lCQT\nGWlyvl9Vh3cFzAXALwMf2nIwyf2q6q5Fnqu6D1X1/AXaPhm4A7hiO/s731OFi3nacHufSPQJRmlM\nnCMjaaeqqs3Ap4HjurTkM0n+DlifZEmX1mxIsjHJGVt+l+QtXRryMeAhQ/s/m+TI7vuJSa5L8qUk\nn0jyU8B/BV7apUHHJnlYkgu6a1yT5Ojut3sn+XiSryT5SxaRSCf5YJcwfTrJs+Yce113D59Ksk+3\n79912xuSXJHkUff5D1TSLs1CRpqwJA8BngZ8jEGx8CTgtVV1CPCfGfz/8qeBxwG/lWRpkiOAI4DH\nAi8FThw6ZQGV5GHAO4HTquqngWdV1Y3AXwNvqKrDq+oyYBVwdne95wJ/1J3ndODLwErgGuARi7id\nM6rqSODpwJ8neWi3/4HAN4CDgS915wY4G3h5d+0/ARxOksYsE/zrXtdOzkpyW5Jrh/b9WZLrk6xL\n8qYkDx46dmaS9d2x4xZzfxYy0uTcP8lVwN8DFwGXdPuvrqqruu+/xKDIWQd8Dngw8HgGhcv7q+pf\nqmoD8Pk55w7wc8DaqloLUFXfmXN8i18EXt315XzgkCT3765xTlXdDbwLWMwqWqck+RRwWdfXld3+\nAs6tqh8B5wEnJnkgcDTwju7ar+v6LKldZ3PP//AC+DiD/1g7isF/9LwCIMkhwBnAkcAzgXOSLFin\nOEdGmpwfVNXhwzsyGLz+2px2r6uq8+a0O5Z7FiPbmlOy2HkmAU6uqq9uoy+LHk1P8kjgBcAJVfWt\nrjjZY8RPlgA/An6uqia31KikqamqS5McOGffJ7Z874bKn95tngys7obfb0iykcF//Fw56homMtJs\n+RhwWjdMRJJHJ3kA8BHgGUke1D2hNDfJKAb/Zz8yyVHdb7c8WXQH8LA513hxkmVdu8O6/R8BTu3+\nC+g5wLIF+ro/8PWuiDmWwVDYFunOtRtwKvCRqroDuBp4QZL7ZeCxi/lDkXQfZIKf7fd8Bik1DP6Z\nsmno2CbggIVOYCIjTc58Kcrw/ncABwJrukeYbweeXlVXdYnHeuAfGRQd9zxR1TeS/DpwXvces00M\nnoy6EPjr7ve/BZwJvA24PskPGAxxvRA4F3gP8H+Bi4EbR91HVX0uyY1JrgeuAz451OZfgZ/o+noj\n8Jpu/3MZzNn5LQZDV6u7e5rvz0fSDLv80ku4/HOXLNxwG5L8PnBHVb1vRLMF/7kw1ZdGSpKk8UlS\nt3733yZ2vZ988NJ7vUexG1q6qKp+Zmjf6QzSmJ+vqju7fb8HUFWv77Y/CvxRVc2dE3gPDi1JkqSJ\nSXIi8DLgpC1FTOdCBg8QLE1yEIOHB9YsdD6HliRJatg0F8RLsho4Htg3yU0Mlnt4BbAU+GT3kMEV\nVfXCqtqQ5GxgLXAXcHotYtjIoSVJkhqVpG77l8kNLe33oHsPLY2biYwkSQ3b1kJ1LXGOjCRJ6i0T\nGUmSWtZ2IGMiI0mS+stERpKkhjUeyJjISJKk/rKQkSRJveXQkiRJDZvmgniTYCIjSZJ6y0RGkqSG\nuSCeJEnSjDKRkSSpYc6RkSRJmlEWMpIkqbcsZCRJUm85R0aSpIY5R0aSJGlGWchIkqTecmhJkqSG\nuSCeJEnSjDKRkSSpYU72lSRJmlEmMpIkNazxQMZERpIk9ZeJjCRJLWs8kjGRkSRJvWUiI0lSw1xH\nRpIkaUaZyEiS1DDXkZEkSZpRFjKSJKm3HFqSJKlhjY8smchIkqT+MpGRJKlljUcyJjKSJKm3TGQk\nSWqYC+JJkiTNKBMZSZIa5oJ4kiRJMypVNe0+SJKkMUgy8X/JV9VEMyALGUmS1FsOLUmSpN6ykJEk\nSb1lISNJknrLQkaSJPWWhYwkSeqt/w/OwzDEl/YJkwAAAABJRU5ErkJggg==\n",
       "text": [
        "<matplotlib.figure.Figure at 0x124da3f50>"
       ]
      }
     ],
     "prompt_number": 23
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "\n",
      "for i in range(len(ltest_list)):\n",
      "    fpr_sk, tpr_sk, thresholds = roc_curve((labels_test==ltest_list[i]), probs_mnb[:,i])\n",
      "    fpr_mnb[ltest_list[i]] = fpr_sk\n",
      "    tpr_mnb[ltest_list[i]] = tpr_sk\n",
      "    try:\n",
      "        auc_mnb[ltest_list[i]] = auc(fpr_sk,tpr_sk)\n",
      "    except:\n",
      "        auc_mnb[ltest_list[i]] = 0\n",
      "    \n",
      "    plt.figure(1,figsize=(10,8))\n",
      "    plt.plot(fpr_mnb[ltest_list[i]],tpr_mnb[ltest_list[i]],label=ltest_list[i])\n",
      "    plt.legend(loc='best')\n",
      "for item in auc_mnb.iteritems():\n",
      "    print item"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "(0, 0.8710911206063976)\n",
        "(1, 0.8710911206063976)\n"
       ]
      },
      {
       "metadata": {},
       "output_type": "display_data",
       "png": "iVBORw0KGgoAAAANSUhEUgAAAlYAAAHfCAYAAACS8paSAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAAAElFTkSuQmCC\n",
       "text": [
        "<matplotlib.figure.Figure at 0x12bf858d0>"
       ]
      }
     ],
     "prompt_number": 24
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [],
     "language": "python",
     "metadata": {},
     "outputs": []
    }
   ],
   "metadata": {}
  }
 ]
}