Notebook Draft.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import nltk\n",
    "import sklearn as sk\n",
    "import pickle\n",
    "\n",
    "train = pickle.load(open('train.p','rb'))\n",
    "test = pickle.load(open('test.p','rb'))\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "toxic severe_toxic --------------\n",
      "n11 965\n",
      "n00 553\n",
      "n10 8272\n",
      "n01 0\n",
      "\n",
      "toxic obscene --------------\n",
      "n11 4797\n",
      "n00 241\n",
      "n10 4440\n",
      "n01 312\n",
      "\n",
      "toxic threat --------------\n",
      "n11 289\n",
      "n00 537\n",
      "n10 8948\n",
      "n01 16\n",
      "\n",
      "toxic insult --------------\n",
      "n11 4445\n",
      "n00 233\n",
      "n10 4792\n",
      "n01 320\n",
      "\n",
      "toxic identity_hate --------------\n",
      "n11 751\n",
      "n00 490\n",
      "n10 8486\n",
      "n01 63\n",
      "\n",
      "severe_toxic obscene --------------\n",
      "n11 921\n",
      "n00 4637\n",
      "n10 44\n",
      "n01 4188\n",
      "\n",
      "severe_toxic threat --------------\n",
      "n11 75\n",
      "n00 8595\n",
      "n10 890\n",
      "n01 230\n",
      "\n",
      "severe_toxic insult --------------\n",
      "n11 833\n",
      "n00 4893\n",
      "n10 132\n",
      "n01 3932\n",
      "\n",
      "severe_toxic identity_hate --------------\n",
      "n11 178\n",
      "n00 8189\n",
      "n10 787\n",
      "n01 636\n",
      "\n",
      "obscene threat --------------\n",
      "n11 198\n",
      "n00 4574\n",
      "n10 4911\n",
      "n01 107\n",
      "\n",
      "obscene insult --------------\n",
      "n11 3739\n",
      "n00 3655\n",
      "n10 1370\n",
      "n01 1026\n",
      "\n",
      "obscene identity_hate --------------\n",
      "n11 612\n",
      "n00 4479\n",
      "n10 4497\n",
      "n01 202\n",
      "\n",
      "threat insult --------------\n",
      "n11 200\n",
      "n00 4920\n",
      "n10 105\n",
      "n01 4565\n",
      "\n",
      "threat identity_hate --------------\n",
      "n11 64\n",
      "n00 8735\n",
      "n10 241\n",
      "n01 750\n",
      "\n",
      "insult identity_hate --------------\n",
      "n11 675\n",
      "n00 4886\n",
      "n10 4090\n",
      "n01 139\n",
      "\n",
      "(('toxic', 'insult'), -0.1356730460565416)\n",
      "(('toxic', 'identity_hate'), -0.10988671131881317)\n",
      "(('toxic', 'obscene'), -0.060333947809904073)\n",
      "(('toxic', 'threat'), 0.01295521313507718)\n",
      "(('threat', 'insult'), 0.044136011520365399)\n",
      "(('obscene', 'threat'), 0.065061945041781657)\n",
      "(('threat', 'identity_hate'), 0.080043377104290964)\n",
      "(('severe_toxic', 'threat'), 0.09188621287898012)\n",
      "(('severe_toxic', 'identity_hate'), 0.12236896326993414)\n",
      "(('severe_toxic', 'insult'), 0.18795553377879667)\n",
      "(('obscene', 'identity_hate'), 0.19199299396789479)\n",
      "(('severe_toxic', 'obscene'), 0.20854274458688682)\n",
      "(('insult', 'identity_hate'), 0.27580983091326855)\n",
      "(('toxic', 'severe_toxic'), 0.32321996943464643)\n",
      "(('obscene', 'insult'), 0.53084469309973004)\n"
     ]
    }
   ],
   "source": [
    "tags = train[train['clean']==False]\n",
    "tags = tags[['toxic', 'severe_toxic', 'obscene', 'threat', 'insult','identity_hate']]\n",
    "#tags['clean'] = tags['clean'].astype(int)\n",
    "\n",
    "def TagCorr(tag1,tag2):\n",
    "    tmp = tags[[tag1,tag2]]\n",
    "    n11 = len(tmp[(tmp[tag1]==1) & (tmp[tag2]==1)])\n",
    "    n10 = len(tmp[(tmp[tag1]==1) & (tmp[tag2]==0)])\n",
    "    n01 = len(tmp[(tmp[tag1]==0) & (tmp[tag2]==1)])\n",
    "    n00 = len(tmp[(tmp[tag1]==0) & (tmp[tag2]==0)])\n",
    "    print(tag1,tag2,'--------------')\n",
    "    print('n11',n11)\n",
    "    print('n00',n00)\n",
    "    print('n10',n10)\n",
    "    print('n01',n01)\n",
    "    print()\n",
    "    \n",
    "    \n",
    "\n",
    "    n1all = n11+n10\n",
    "    n0all = n00+n01\n",
    "    nall1 = n11+n01\n",
    "    nall0 = n00+n01\n",
    "\n",
    "    phi = (n11*n00 - n01*n10)/(np.sqrt(nall0*nall1*n0all*n1all))\n",
    "    return phi\n",
    "from itertools import combinations\n",
    "scores = {}\n",
    "for item in combinations(['toxic', 'severe_toxic', 'obscene', 'threat', 'insult','identity_hate'],2):\n",
    "    col1=item[0]\n",
    "    col2=item[1]\n",
    "    #print(tags.pivot(index=[col1],columns=[col2]))\n",
    "    #print(item, '-------- ')\n",
    "    scores[(col1,col2)] = TagCorr(col1,col2)\n",
    "# #print(scores)\n",
    "import operator\n",
    "x = scores\n",
    "sorted_x = sorted(x.items(), key=operator.itemgetter(1))\n",
    "for ii in sorted_x:\n",
    "    print(ii)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Index(['comment_text', 'toxic', 'severe_toxic', 'obscene', 'threat', 'insult',\n",
      "       'identity_hate', 'clean'],\n",
      "      dtype='object')\n",
      "toxic (9237, 8)\n",
      "severe_toxic (965, 8)\n",
      "obscene (5109, 8)\n",
      "threat (305, 8)\n",
      "insult (4765, 8)\n",
      "identity_hate (814, 8)\n",
      "clean (86061, 8)\n"
     ]
    }
   ],
   "source": [
    "print(train.columns)\n",
    "for item in ['toxic', 'severe_toxic', 'obscene', 'threat', 'insult','identity_hate','clean']:\n",
    "    print(item,train[train[item] == 1].shape)\n",
    "# train['comment_text'] = train['comment_text'].str.lower()\n",
    "# train['comment_text'] = train['comment_text'].str.replace('^\\d+|\\n|\\t|\"|==|;|:|@', '', case=False)\n",
    "# train['comment_text'] = train['comment_text'].str.replace('\\d+', '')\n",
    "# train['comment_text'] = train['comment_text'].str.replace(',', '')\n",
    "# train['comment_text'] = train['comment_text'].str.replace('?', '')\n",
    "# train['comment_text'] = train['comment_text'].str.replace('!', '')\n",
    "# train['comment_text'] = train['comment_text'].str.replace('&', '')\n",
    "# train['comment_text'] = train['comment_text'].str.replace('[...]', '')\n",
    "# train['comment_text'] = train['comment_text'].str.replace('[....]', '')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "clean = train.loc[np.random.choice(train[train['clean']==True].index.tolist(),size=9237)].reset_index(drop=True)\n",
    "severe_toxic = train.loc[np.random.choice(train[train['severe_toxic']==True].index.tolist(),size=6000,replace=True)].reset_index(drop=True)\n",
    "obscene = train.loc[np.random.choice(train[train['obscene']==True].index.tolist(),size=2000,replace=True)].reset_index(drop=True)\n",
    "threat = train.loc[np.random.choice(train[train['threat']==True].index.tolist(),size=6000,replace=True)].reset_index(drop=True)\n",
    "insult = train.loc[np.random.choice(train[train['insult']==True].index.tolist(),size=2300,replace=True)].reset_index(drop=True)\n",
    "identity_hate = train.loc[np.random.choice(train[train['identity_hate']==True].index.tolist(),size=6000,replace=True)].reset_index(drop=True)\n",
    "toxic = train[train['toxic']==1].reset_index(drop=True)\n",
    "\n",
    "# clean = train.loc[np.random.choice(train[train['clean']==True].index.tolist(),size=300)].reset_index(drop=True)\n",
    "# toxic = train.loc[np.random.choice(train[train['toxic']==True].index.tolist(),size=300,replace=False)].reset_index(drop=True)\n",
    "# obscene = train.loc[np.random.choice(train[train['obscene']==True].index.tolist(),size=300,replace=False)].reset_index(drop=True)\n",
    "# severe_toxic = train.loc[np.random.choice(train[train['severe_toxic']==True].index.tolist(),size=300,replace=True)].reset_index(drop=True)\n",
    "# insult = train.loc[np.random.choice(train[train['insult']==True].index.tolist(),size=300,replace=False)].reset_index(drop=True)\n",
    "# identity_hate = train.loc[np.random.choice(train[train['identity_hate']==True].index.tolist(),size=300,replace=True)].reset_index(drop=True)\n",
    "# threat = train[train['threat']==1].reset_index(drop=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(40774, 8)\n"
     ]
    }
   ],
   "source": [
    "sampled_train = pd.concat([clean,threat,toxic,insult,identity_hate,obscene,severe_toxic]).reset_index(drop=True)\n",
    "#sampled_train = sampled_train.sample(frac=1).reset_index(drop=True)\n",
    "sampled_train['clean'] = sampled_train['clean'].astype(int)\n",
    "print(sampled_train.shape)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "300.7142857142857"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# from sklearn.decomposition import SparsePCA\n",
    "# from sklearn.decomposition import TruncatedSVD\n",
    "# from scipy import sparse as sp\n",
    "\n",
    "# pca = TruncatedSVD(100)\n",
    "# X = pca.fit_transform(feature_matrix)\n",
    "2105 /7\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2105, 120)\n",
      "tSNE starts ...\n"
     ]
    },
    {
     "ename": "IndexError",
     "evalue": "index 2 is out of bounds for axis 1 with size 2",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mIndexError\u001b[0m                                Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-6-4b01972d11a2>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[0;32m     16\u001b[0m             \u001b[0mplt\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mplot\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mX_embedded\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mii\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m:\u001b[0m\u001b[0mii\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mX_embedded\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mii\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m:\u001b[0m\u001b[0mii\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mii\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m2\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0malpha\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m.6\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     17\u001b[0m             \u001b[0mplt\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msubplot\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;36m122\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 18\u001b[1;33m             \u001b[0mplt\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mplot\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mX_embedded\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mii\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m:\u001b[0m\u001b[0mii\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mX_embedded\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mii\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m:\u001b[0m\u001b[0mii\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m2\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mii\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m2\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0malpha\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m.6\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     19\u001b[0m             \u001b[0mplt\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msubplot\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;36m123\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     20\u001b[0m             \u001b[0mplt\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mplot\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mX_embedded\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mii\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m:\u001b[0m\u001b[0mii\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mX_embedded\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mii\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m:\u001b[0m\u001b[0mii\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m2\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mii\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m2\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0malpha\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m.6\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mIndexError\u001b[0m: index 2 is out of bounds for axis 1 with size 2"
     ]
    }
   ],
   "source": [
    "# plt.plot(pca.explained_variance_ratio_,'*' )\n",
    "# plt.show()\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "sampled_train.shape\n",
    "from sklearn.manifold import TSNE\n",
    "feature_model = TfidfVectorizer(ngram_range=(1,4),max_features=120)\n",
    "feature_matrix = feature_model.fit_transform(sampled_train['comment_text'])\n",
    "print(feature_matrix.todense().shape)\n",
    "print('tSNE starts ...')\n",
    "for lr in [1000]:\n",
    "    for prep in [1,10,50,100,150]:\n",
    "        X_embedded = TSNE(n_components=2,learning_rate =lr,perplexity =prep).fit_transform(feature_matrix.todense())\n",
    "        plt.figure()#figsize=(12,12))\n",
    "        for ii in [(0,300,'*'),(300,600,'r*'),(600,900,'gs'),(900,1200,'ko'),(1200,1500,'cs'),(1500,1800,'k^'),(1800,2104,'r^')]:\n",
    "            plt.subplot(121)\n",
    "            plt.plot(X_embedded[ii[0]:ii[1],0],X_embedded[ii[0]:ii[1],1],ii[2],alpha=.6)\n",
    "            plt.subplot(122)\n",
    "            plt.plot(X_embedded[ii[0]:ii[1],0],X_embedded[ii[0]:ii[1],2],ii[2],alpha=.6)\n",
    "            plt.subplot(123)\n",
    "            plt.plot(X_embedded[ii[0]:ii[1],1],X_embedded[ii[0]:ii[1],2],ii[2],alpha=.6)\n",
    "        plt.title('prep='+str(prep)+' itr='+str(lr))\n",
    "        plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(226997,)"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "toxic\n",
      "severe_toxic\n",
      "obscene\n",
      "threat\n",
      "insult\n",
      "identity_hate\n"
     ]
    }
   ],
   "source": [
    "from sklearn.naive_bayes import MultinomialNB\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer, CountVectorizer\n",
    "from sklearn.metrics import classification_report\n",
    "\n",
    "\n",
    "# x_fit,x_valid,y_fit,y_valid = train_test_split(sampled_train['comment_text'],sampled_train[['toxic', 'severe_toxic', 'obscene', 'threat', 'insult','identity_hate','clean']],test_size=.5,random_state=46)\n",
    "# x_fit = feature_model.fit_transform(x_fit)\n",
    "# x_valid = feature_model.transform(x_valid)\n",
    "#sample = pd.read_csv('sample_submission.csv')\n",
    "sample_tfidf = pd.read_csv('sample_submission.csv')\n",
    "\n",
    "feature_model_tfidf = TfidfVectorizer(ngram_range=(1,4),max_features=250000)\n",
    "feature_matrix_tfidf = feature_model_tfidf.fit_transform(sampled_train['comment_text'])\n",
    "test_x = feature_model_tfidf.transform(test)\n",
    "for col in ['toxic', 'severe_toxic', 'obscene', 'threat', 'insult','identity_hate']:\n",
    "#     x_fit,x_valid,y_fit,y_valid = train_test_split(sampled_train['comment_text'],sampled_train[col],test_size=.5,random_state=46)\n",
    "#     feature_model = TfidfVectorizer(ngram_range=(1,4))\n",
    "#     x_fit = feature_model.fit_transform(x_fit)\n",
    "#     x_valid = feature_model.transform(x_valid)\n",
    "\n",
    "#     feature_model_tfidf = TfidfVectorizer(ngram_range=(1,4),max_features=250000)\n",
    "#     feature_matrix_tfidf = feature_model_tfidf.fit_transform(sampled_train['comment_text'])\n",
    "#     test_x = feature_model_tfidf.transform(test)\n",
    "\n",
    "    clf = MultinomialNB()\n",
    "    clf.fit(feature_matrix_tfidf,sampled_train[col])\n",
    "    preds = clf.predict_proba(test_x)[:,1]\n",
    "    preds = np.insert(preds,52300,0.5)\n",
    "    print(col)\n",
    "    #print(preds)\n",
    "    #print(sample.shape,preds.shape)\n",
    "    #print(classification_report(y_valid,preds))\n",
    "    sample_tfidf[col] = preds\n",
    "\n",
    "sample_tfidf.to_csv(\"result_tfidf_250k.csv\", index=False)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(226998, 7)"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sample_tfidf.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "id               1.237033e+11\n",
      "toxic            4.993011e-01\n",
      "severe_toxic     9.112207e-01\n",
      "obscene          6.409270e-01\n",
      "threat           4.628263e-02\n",
      "insult           6.410021e-01\n",
      "identity_hate    4.331520e-02\n",
      "Name: 27991, dtype: float64\n",
      "id               2.051069e+11\n",
      "toxic            8.201353e-05\n",
      "severe_toxic     9.895908e-01\n",
      "obscene          2.414067e-03\n",
      "threat           1.181419e-02\n",
      "insult           2.311675e-03\n",
      "identity_hate    2.636068e-03\n",
      "Name: 46355, dtype: float64\n",
      "id               2.266650e+11\n",
      "toxic            4.415038e-01\n",
      "severe_toxic     9.545650e-01\n",
      "obscene          6.493189e-01\n",
      "threat           5.387702e-02\n",
      "insult           6.512209e-01\n",
      "identity_hate    4.157116e-02\n",
      "Name: 51240, dtype: float64\n",
      "id               3.228455e+11\n",
      "toxic            2.830413e-01\n",
      "severe_toxic     7.503823e-01\n",
      "obscene          3.538911e-01\n",
      "threat           1.011477e-02\n",
      "insult           3.756172e-01\n",
      "identity_hate    8.623781e-03\n",
      "Name: 73023, dtype: float64\n",
      "id               3.276917e+11\n",
      "toxic            2.947374e-01\n",
      "severe_toxic     5.925070e-01\n",
      "obscene          2.994375e-01\n",
      "threat           1.376234e-02\n",
      "insult           3.258210e-01\n",
      "identity_hate    9.902037e-03\n",
      "Name: 74143, dtype: float64\n",
      "id               4.429626e+11\n",
      "toxic            2.934033e-01\n",
      "severe_toxic     8.130656e-01\n",
      "obscene          4.125320e-01\n",
      "threat           1.257177e-02\n",
      "insult           3.879354e-01\n",
      "identity_hate    7.947873e-03\n",
      "Name: 100416, dtype: float64\n",
      "id               4.885325e+11\n",
      "toxic            4.529707e-01\n",
      "severe_toxic     5.855053e-01\n",
      "obscene          4.100226e-01\n",
      "threat           6.172710e-02\n",
      "insult           4.103051e-01\n",
      "identity_hate    6.077212e-02\n",
      "Name: 110662, dtype: float64\n",
      "id               6.305018e+11\n",
      "toxic            3.402262e-01\n",
      "severe_toxic     7.377534e-01\n",
      "obscene          3.780110e-01\n",
      "threat           3.203957e-02\n",
      "insult           3.530139e-01\n",
      "identity_hate    1.850295e-02\n",
      "Name: 143062, dtype: float64\n",
      "id               6.516194e+11\n",
      "toxic            3.279508e-01\n",
      "severe_toxic     6.710390e-01\n",
      "obscene          3.580570e-01\n",
      "threat           6.888183e-03\n",
      "insult           3.706734e-01\n",
      "identity_hate    1.193162e-02\n",
      "Name: 147862, dtype: float64\n",
      "id               7.789989e+11\n",
      "toxic            4.925318e-01\n",
      "severe_toxic     8.401441e-01\n",
      "obscene          5.532289e-01\n",
      "threat           4.635098e-02\n",
      "insult           5.457670e-01\n",
      "identity_hate    3.820414e-02\n",
      "Name: 176728, dtype: float64\n",
      "id               8.324822e+11\n",
      "toxic            3.955216e-01\n",
      "severe_toxic     9.092826e-01\n",
      "obscene          5.397106e-01\n",
      "threat           3.729489e-02\n",
      "insult           5.420755e-01\n",
      "identity_hate    1.903104e-02\n",
      "Name: 188949, dtype: float64\n",
      "id               9.318392e+11\n",
      "toxic            4.241411e-01\n",
      "severe_toxic     7.725180e-01\n",
      "obscene          3.869010e-01\n",
      "threat           1.635190e-02\n",
      "insult           4.246046e-01\n",
      "identity_hate    2.199851e-02\n",
      "Name: 211411, dtype: float64\n",
      "id               9.427736e+11\n",
      "toxic            4.742519e-01\n",
      "severe_toxic     8.759864e-01\n",
      "obscene          5.357593e-01\n",
      "threat           1.600690e-02\n",
      "insult           5.158610e-01\n",
      "identity_hate    2.206583e-02\n",
      "Name: 213920, dtype: float64\n",
      "id               9.810943e+11\n",
      "toxic            4.225272e-01\n",
      "severe_toxic     7.280405e-01\n",
      "obscene          4.558929e-01\n",
      "threat           2.832557e-02\n",
      "insult           4.664793e-01\n",
      "identity_hate    3.266033e-02\n",
      "Name: 222713, dtype: float64\n"
     ]
    }
   ],
   "source": [
    "ii = 0\n",
    "for item in sample_tfidf.index:\n",
    "    if (sample_tfidf.loc[item,'toxic']<.5) and (sample_tfidf.loc[item,'severe_toxic']>.5):# or (sample_tfidf.loc[item,'obscene']>.5) or (sample_tfidf.loc[item,'threat']>.5) or (sample_tfidf.loc[item,'insult']>.5)):\n",
    "        ii += 1\n",
    "        print(sample_tfidf.loc[item])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "14.771495784103825"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "3353100/226998"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(226997,)\n",
      "(226998,)\n"
     ]
    }
   ],
   "source": [
    "tmp = preds[:,1]\n",
    "print(tmp.shape)\n",
    "tmp = np.insert(tmp,52300,0.5)\n",
    "print(tmp.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{231298963278}"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "set(sample['id'].tolist())-set(test.index.tolist())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "52300"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sample['id'].tolist().index(231298963278)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# train['comment_text'] = train['comment_text'].str.lower()\n",
    "# train['comment_text'] = train['comment_text'].str.replace('^\\d+|\\n|\\t|\"|==|;|:|@', '', case=False)\n",
    "# train['comment_text'] = train['comment_text'].str.replace('\\d+', '')\n",
    "# train['comment_text'] = train['comment_text'].str.replace(',', '')\n",
    "# train['comment_text'] = train['comment_text'].str.replace('?', '')\n",
    "# train['comment_text'] = train['comment_text'].str.replace('!', '')\n",
    "# train['comment_text'] = train['comment_text'].str.replace('&', '')\n",
    "# train['comment_text'] = train['comment_text'].str.replace('[...]', '')\n",
    "# train['comment_text'] = train['comment_text'].str.replace('[....]', '')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "toxic = train[train['toxic']==1]\n",
    "severe_toxic = train[train['severe_toxic']==1]\n",
    "obscene = train[train['obscene']==1]\n",
    "threat = train[train['threat']==1]\n",
    "insult = train[train['insult']==1]\n",
    "identity_hate = train[train['identity_hate']==1]\n",
    "clean\n",
    "print(toxic.head().index.tolist())\n",
    "print(nltk.word_tokenize(toxic.loc[0]['comment_text']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from collections import Counter\n",
    "def GetBestWords(df,n):\n",
    "    tmp = []\n",
    "    c = Counter(tmp)\n",
    "    for ii in df.index.tolist():\n",
    "        c.update(Counter(nltk.word_tokenize(df.loc[ii]['comment_text'])))\n",
    "    \n",
    "    return c.most_common(n)\n",
    "\n",
    "\n",
    "print(GetBestWords(toxic,10))\n",
    "print(GetBestWords(severe_toxic,10))\n",
    "print(GetBestWords(obscene,10))\n",
    "print(GetBestWords(threat,10))\n",
    "print(GetBestWords(insult,10))\n",
    "print(GetBestWords(identity_hate,10))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sk.feature_extraction.text import CountVectorizer\n",
    "    \n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "tmp = train[['comment_text','toxic']]\n",
    "data_new = tmp[tmp['toxic'] == 1]\n",
    "inds = np.random.choice(data_new.index.tolist(),size=5)\n",
    "print(inds)\n",
    "data_new_new = data_new.loc[inds]\n",
    "data_new_new\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "lbls = np.array(train[train.columns.tolist()[2::]]).tolist()\n",
    "tmp = [''.join([str(jj) for jj in ii]) for ii in lbls]\n",
    "print(tmp[0])\n",
    "print(Counter(tmp))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import pickle\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import nltk\n",
    "import sklearn as sk"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "x_train = pickle.load(open('x_train.p','rb'))\n",
    "y_train = pickle.load(open('y_train.p','rb'))\n",
    "y_trainNew = pickle.load(open('y_trainNew.p','rb'))\n",
    "train = pickle.load(open('train.p','rb'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.naive_bayes import MultinomialNB\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "from sklearn.metrics import classification_report\n",
    "\n",
    "\n",
    "feature_model = TfidfVectorizer()\n",
    "\n",
    "x_fit,x_valid,y_fit,y_valid = train_test_split(x_train,y_trainNew,test_size=.3,random_state=46)\n",
    "x_fit = feature_model.fit_transform(x_fit)\n",
    "\n",
    "\n",
    "clf = MultinomialNB()\n",
    "clf.fit(x_fit,y_fit)\n",
    "preds = clf.predict(feature_model.transform(x_valid))\n",
    "print(classification_report(y_valid,preds))\n",
    "#clean --> True\n",
    "#toxic --> False"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "print(preds.tolist().count(True))\n",
    "print(preds.tolist().count(False))\n",
    "print(train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "x = feature_model.transform(train['comment_text'])\n",
    "preds_global = clf.predict(x)\n",
    "print(classification_report(train['clean'],preds_global))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "binary_clf = clf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "toxic = train[train['clean']==False]\n",
    "toxic.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "print(len(toxic[toxic['toxic']==1]))\n",
    "print(len(toxic[toxic['severe_toxic']==1]))\n",
    "print(len(toxic[toxic['obscene']==1]))\n",
    "print(len(toxic[toxic['identity_hate']==1]))\n",
    "print(len(toxic[toxic['insult']==1]))\n",
    "print(len(toxic[toxic['threat']==1]))\n",
    "toxic['toxic']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "test = pickle.load(open('test.p','rb'))\n",
    "test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "x_Test =  feature_model.transform(test)\n",
    "pred_clean = binary_clf.predict_proba(x_Test)\n",
    "print(pred_clean)\n",
    "print(len([ii for ii in pred_clean.tolist() if ii[1]>ii[0]]))\n",
    "final = np.zeros((226997,6))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "for col in ['toxic', 'severe_toxic', 'obscene', 'threat', 'insult','identity_hate']:\n",
    "    #x_fit,x_valid,y_fit,y_valid = train_test_split(toxic['comment_text'],toxic[col],test_size=.5,random_state=46)\n",
    "    toxic_feature_matrix = TfidfVectorizer().fit_transform(toxic['comment_text'])\n",
    "    toxic_clf = MultinomialNB()\n",
    "    toxic_clf.fit(toxic_feature_matrix,toxic[col])\n",
    "    pred = toxic_clf.predict(toxic_feature_model.transform(x_valid))\n",
    "    print(col,classification_report(y_valid,pred))\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "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>comment_text</th>\n",
       "      <th>toxic</th>\n",
       "      <th>severe_toxic</th>\n",
       "      <th>obscene</th>\n",
       "      <th>threat</th>\n",
       "      <th>insult</th>\n",
       "      <th>identity_hate</th>\n",
       "      <th>clean</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>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",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>22256635</th>\n",
       "      <td>nonsense kiss off geek what i say be true ill ...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27450690</th>\n",
       "      <td>please do not vandalize page as -PRON- do with...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>54037174</th>\n",
       "      <td>point of interest i remove the point of intere...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>77493077</th>\n",
       "      <td>ask some -PRON- nationality be a racial offenc...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>79357270</th>\n",
       "      <td>the reader here be not go by -PRON- say so for...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>82428052</th>\n",
       "      <td>fry chicken be dat sum fry chicken</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>87311443</th>\n",
       "      <td>why can -PRON- put english for example on some...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>114749757</th>\n",
       "      <td>guy fawk -PRON- be a resident in bridgwater an...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>138560519</th>\n",
       "      <td>as far as nickname go this article be embarras...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>139353149</th>\n",
       "      <td>woodland meadow good to hear that -PRON- corre...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>150009866</th>\n",
       "      <td>well i just finish a good bit of editing i hav...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>152276337</th>\n",
       "      <td>discussion should take place on the article ta...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>153331729</th>\n",
       "      <td>uh oh -PRON- call -PRON- bluff i be intimidate...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>153600803</th>\n",
       "      <td>-PRON- should also contact the live descendant...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>154186883</th>\n",
       "      <td>may utc notability of -PRON- new heart a tag h...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>161894108</th>\n",
       "      <td>while i agree that this article be not fa mate...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>169740962</th>\n",
       "      <td>a turkish citizen and -PRON- have receive an a...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>178084608</th>\n",
       "      <td>please explain why censorship of quality addit...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>208005265</th>\n",
       "      <td>in any case this edit war will last forever -P...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>224733383</th>\n",
       "      <td>vandalism of george washington who be -PRON- t...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>225701312</th>\n",
       "      <td>why have not alitalia be remove rom the allian...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>237279459</th>\n",
       "      <td>another afd stat example i hope -PRON- do not ...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>254015372</th>\n",
       "      <td>i will how about ah -PRON- have get nothing to...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>280570905</th>\n",
       "      <td>i have move some tedious detail in survey of h...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>284253328</th>\n",
       "      <td>anniehall what separate this from capitalism b...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>287371884</th>\n",
       "      <td>and -PRON- also not random -PRON- be the first...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>293668009</th>\n",
       "      <td>the graceful slick be non other than an ungrac...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>300891545</th>\n",
       "      <td>regard edit make during december utc thank -PR...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>307040060</th>\n",
       "      <td>the section be now call discrepancy and inaccu...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>340940431</th>\n",
       "      <td>smackdown good smackdown on qatar -PRON- be no...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999674711016</th>\n",
       "      <td>god -PRON- be an anglonormanist really not an ...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999677373048</th>\n",
       "      <td>t cotton letter to the ayatollah of iran dwp i...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999693619845</th>\n",
       "      <td>much appreciate thank -PRON- yours be -PRON- f...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999714130533</th>\n",
       "      <td>i do apologize once more to -PRON- for -PRON- ...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999716611707</th>\n",
       "      <td>-PRON- have be confirm that raul have join fc ...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999731184003</th>\n",
       "      <td>that an article be more than a definition do n...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999741263322</th>\n",
       "      <td>call someone the archetypical unscientific eco...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999744390044</th>\n",
       "      <td>just point out that if -PRON- intend to keep a...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999749161210</th>\n",
       "      <td>-PRON- be sorry for not notice that on the art...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999776197636</th>\n",
       "      <td>can i just say noone care about -PRON- opinion...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999781781533</th>\n",
       "      <td>see wikipediaadministratorsmisuseofadministrat...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999825778693</th>\n",
       "      <td>december please stop the foolish edit to the a...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999829405247</th>\n",
       "      <td>another unblock request</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999856640855</th>\n",
       "      <td>so this can finally be over with</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999883234532</th>\n",
       "      <td>oh okay fair enough then thank for make the ef...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999887342595</th>\n",
       "      <td>i believe -PRON- be out of line -PRON- be spec...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999888096837</th>\n",
       "      <td>april utc secondhander</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999898414104</th>\n",
       "      <td>-PRON- stay let move on corbett</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999903896729</th>\n",
       "      <td>conflict of interest pmshef this be an attempt...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999909788460</th>\n",
       "      <td>perhaps the single most potent way to balance ...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999912713635</th>\n",
       "      <td>please do not bother i be just wonder also i b...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999939579242</th>\n",
       "      <td>the article the eighth sea have be speedily de...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999945355747</th>\n",
       "      <td>each alum agree to how much information can be...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999950278781</th>\n",
       "      <td>caucasion vs white i notice that there be a ch...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999976306590</th>\n",
       "      <td>this culture allow people to hold -PRON- wife ...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999977655955</th>\n",
       "      <td>i have discuss -PRON- unlike most of those who...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999982426659</th>\n",
       "      <td>ps almost forget paine do not reply back to th...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999982764066</th>\n",
       "      <td>mamoun darkazanli for some reason i be unable ...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999986890563</th>\n",
       "      <td>salafi would be a good term -PRON- be more pol...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999988164717</th>\n",
       "      <td>make wikipedia a good and more inviting place</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>95851 rows × 8 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                   comment_text  toxic  \\\n",
       "id                                                                       \n",
       "22256635      nonsense kiss off geek what i say be true ill ...      1   \n",
       "27450690      please do not vandalize page as -PRON- do with...      0   \n",
       "54037174      point of interest i remove the point of intere...      0   \n",
       "77493077      ask some -PRON- nationality be a racial offenc...      0   \n",
       "79357270      the reader here be not go by -PRON- say so for...      0   \n",
       "82428052                     fry chicken be dat sum fry chicken      0   \n",
       "87311443      why can -PRON- put english for example on some...      0   \n",
       "114749757     guy fawk -PRON- be a resident in bridgwater an...      0   \n",
       "138560519     as far as nickname go this article be embarras...      0   \n",
       "139353149     woodland meadow good to hear that -PRON- corre...      0   \n",
       "150009866     well i just finish a good bit of editing i hav...      0   \n",
       "152276337     discussion should take place on the article ta...      0   \n",
       "153331729     uh oh -PRON- call -PRON- bluff i be intimidate...      0   \n",
       "153600803     -PRON- should also contact the live descendant...      0   \n",
       "154186883     may utc notability of -PRON- new heart a tag h...      0   \n",
       "161894108     while i agree that this article be not fa mate...      0   \n",
       "169740962     a turkish citizen and -PRON- have receive an a...      0   \n",
       "178084608     please explain why censorship of quality addit...      0   \n",
       "208005265     in any case this edit war will last forever -P...      0   \n",
       "224733383     vandalism of george washington who be -PRON- t...      0   \n",
       "225701312     why have not alitalia be remove rom the allian...      1   \n",
       "237279459     another afd stat example i hope -PRON- do not ...      0   \n",
       "254015372     i will how about ah -PRON- have get nothing to...      0   \n",
       "280570905     i have move some tedious detail in survey of h...      0   \n",
       "284253328     anniehall what separate this from capitalism b...      0   \n",
       "287371884     and -PRON- also not random -PRON- be the first...      0   \n",
       "293668009     the graceful slick be non other than an ungrac...      1   \n",
       "300891545     regard edit make during december utc thank -PR...      0   \n",
       "307040060     the section be now call discrepancy and inaccu...      0   \n",
       "340940431     smackdown good smackdown on qatar -PRON- be no...      0   \n",
       "...                                                         ...    ...   \n",
       "999674711016  god -PRON- be an anglonormanist really not an ...      0   \n",
       "999677373048  t cotton letter to the ayatollah of iran dwp i...      0   \n",
       "999693619845  much appreciate thank -PRON- yours be -PRON- f...      0   \n",
       "999714130533  i do apologize once more to -PRON- for -PRON- ...      0   \n",
       "999716611707  -PRON- have be confirm that raul have join fc ...      0   \n",
       "999731184003  that an article be more than a definition do n...      0   \n",
       "999741263322  call someone the archetypical unscientific eco...      0   \n",
       "999744390044  just point out that if -PRON- intend to keep a...      0   \n",
       "999749161210  -PRON- be sorry for not notice that on the art...      0   \n",
       "999776197636  can i just say noone care about -PRON- opinion...      0   \n",
       "999781781533  see wikipediaadministratorsmisuseofadministrat...      0   \n",
       "999825778693  december please stop the foolish edit to the a...      0   \n",
       "999829405247                            another unblock request      0   \n",
       "999856640855                   so this can finally be over with      0   \n",
       "999883234532  oh okay fair enough then thank for make the ef...      0   \n",
       "999887342595  i believe -PRON- be out of line -PRON- be spec...      0   \n",
       "999888096837                             april utc secondhander      0   \n",
       "999898414104                    -PRON- stay let move on corbett      0   \n",
       "999903896729  conflict of interest pmshef this be an attempt...      0   \n",
       "999909788460  perhaps the single most potent way to balance ...      0   \n",
       "999912713635  please do not bother i be just wonder also i b...      0   \n",
       "999939579242  the article the eighth sea have be speedily de...      0   \n",
       "999945355747  each alum agree to how much information can be...      0   \n",
       "999950278781  caucasion vs white i notice that there be a ch...      0   \n",
       "999976306590  this culture allow people to hold -PRON- wife ...      1   \n",
       "999977655955  i have discuss -PRON- unlike most of those who...      0   \n",
       "999982426659  ps almost forget paine do not reply back to th...      1   \n",
       "999982764066  mamoun darkazanli for some reason i be unable ...      0   \n",
       "999986890563  salafi would be a good term -PRON- be more pol...      0   \n",
       "999988164717      make wikipedia a good and more inviting place      0   \n",
       "\n",
       "              severe_toxic  obscene  threat  insult  identity_hate  clean  \n",
       "id                                                                         \n",
       "22256635                 0        0       0       0              0  False  \n",
       "27450690                 0        0       0       0              0   True  \n",
       "54037174                 0        0       0       0              0   True  \n",
       "77493077                 0        0       0       0              0   True  \n",
       "79357270                 0        0       0       0              0   True  \n",
       "82428052                 0        0       0       0              0   True  \n",
       "87311443                 0        0       0       0              0   True  \n",
       "114749757                0        0       0       0              0   True  \n",
       "138560519                0        0       0       0              0   True  \n",
       "139353149                0        0       0       0              0   True  \n",
       "150009866                0        0       0       0              0   True  \n",
       "152276337                0        0       0       0              0   True  \n",
       "153331729                0        0       0       0              0   True  \n",
       "153600803                0        0       0       0              0   True  \n",
       "154186883                0        0       0       0              0   True  \n",
       "161894108                0        0       0       0              0   True  \n",
       "169740962                0        0       0       0              0   True  \n",
       "178084608                0        0       0       0              0   True  \n",
       "208005265                0        0       0       0              0   True  \n",
       "224733383                0        0       0       0              0   True  \n",
       "225701312                0        0       0       0              0  False  \n",
       "237279459                0        0       0       0              0   True  \n",
       "254015372                0        0       0       0              0   True  \n",
       "280570905                0        0       0       0              0   True  \n",
       "284253328                0        0       0       0              0   True  \n",
       "287371884                0        0       0       0              0   True  \n",
       "293668009                0        0       0       0              0  False  \n",
       "300891545                0        0       0       0              0   True  \n",
       "307040060                0        0       0       0              0   True  \n",
       "340940431                0        0       0       0              0   True  \n",
       "...                    ...      ...     ...     ...            ...    ...  \n",
       "999674711016             0        0       0       0              0   True  \n",
       "999677373048             0        0       0       0              0   True  \n",
       "999693619845             0        0       0       0              0   True  \n",
       "999714130533             0        0       0       0              0   True  \n",
       "999716611707             0        0       0       0              0   True  \n",
       "999731184003             0        0       0       0              0   True  \n",
       "999741263322             0        0       0       0              0   True  \n",
       "999744390044             0        0       0       0              0   True  \n",
       "999749161210             0        0       0       0              0   True  \n",
       "999776197636             0        0       0       0              0   True  \n",
       "999781781533             0        0       0       0              0   True  \n",
       "999825778693             0        0       0       0              0   True  \n",
       "999829405247             0        0       0       0              0   True  \n",
       "999856640855             0        0       0       0              0   True  \n",
       "999883234532             0        0       0       0              0   True  \n",
       "999887342595             0        0       0       0              0   True  \n",
       "999888096837             0        0       0       0              0   True  \n",
       "999898414104             0        0       0       0              0   True  \n",
       "999903896729             0        0       0       0              0   True  \n",
       "999909788460             0        0       0       0              0   True  \n",
       "999912713635             0        0       0       0              0   True  \n",
       "999939579242             0        0       0       0              0   True  \n",
       "999945355747             0        0       0       0              0   True  \n",
       "999950278781             0        0       0       0              0   True  \n",
       "999976306590             0        1       0       0              0  False  \n",
       "999977655955             0        0       0       0              0   True  \n",
       "999982426659             0        1       0       0              0  False  \n",
       "999982764066             0        0       0       0              0   True  \n",
       "999986890563             0        0       0       0              0   True  \n",
       "999988164717             0        0       0       0              0   True  \n",
       "\n",
       "[95851 rows x 8 columns]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pickle\n",
    "train = pickle.load(open('train.p','rb'))\n",
    "train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "ename": "AttributeError",
     "evalue": "'Counter' object has no attribute 'iteritems'",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mAttributeError\u001b[0m                            Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-28-8e3e72a296d3>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[0;32m      6\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      7\u001b[0m \u001b[1;31m#print(words)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 8\u001b[1;33m \u001b[0mprint\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0msum\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mwords\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0miteritems\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;31mAttributeError\u001b[0m: 'Counter' object has no attribute 'iteritems'"
     ]
    }
   ],
   "source": [
    "from collections import Counter\n",
    "words = Counter()\n",
    "\n",
    "for item in train[(train['insult']==1) & (train['toxic']==0) & (train['severe_toxic']==0) & (train['identity_hate']==0) & (train['threat']==0)]['comment_text'].tolist():\n",
    "    words.update(item.split(\" \"))\n",
    "    \n",
    "#print(words)\n",
    "print(sum(words.iteritems()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAtoAAAK9CAYAAAAaH8vRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x19d1285b0f0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 0.34358928  0.2647242   0.22857357  0.20052353  0.15362363]\n"
     ]
    }
   ],
   "source": [
    "TfIDF = [[ 0.25973704,  0.38259231,  0.35260526,  0.25524401,  0.42347124,  0.3878858 ],\n",
    " [ 0.22764892,  0.31051723,  0.30553084,  0.1174998,   0.35793688,  0.26921151],\n",
    " [ 0.22136163,  0.2459898,   0.287604,    0.04395131,  0.32820432,  0.24433037],\n",
    " [ 0.20873702,  0.20027658,  0.27754501,  0.01976685,  0.31014506,  0.18667069],\n",
    " [ 0.21508665,  0.11841287,  0.22807661,  0.0191923,   0.25326522,  0.08770811]]\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import nltk\n",
    "import sklearn as sk\n",
    "import pickle\n",
    "TfIDF = np.array(TfIDF)\n",
    "plt.figure(figsize=(12,12))\n",
    "plt.xlabel('Number of Features')\n",
    "plt.ylabel('Log Loss Error')\n",
    "cols = ['toxic','severe','obscene','threat','insult','identity']\n",
    "for ii in range(6):\n",
    "    plt.plot([1000,10000,50000,100000,250000],TfIDF[:,ii],'-s',label=cols[ii])\n",
    "plt.legend()\n",
    "plt.show()\n",
    "print(np.mean(TfIDF,axis=1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "Count  [[ 5.44005352  3.24442325  3.48384799  1.25401956  3.81970083  2.05564242]\n",
    " [ 5.36710532  3.11122999  3.34588574  0.41492294  3.54279776  1.15942275]\n",
    " [ 4.79755802  1.93839196  2.909193    0.15455915  3.09133384  0.80558492]\n",
    " [ 4.24056479  1.2191048   2.57906646  0.11906803  2.69174137  0.52467048]\n",
    " [ 3.32067058  0.54886593  1.81057233  0.11293797  1.8988395   0.22814458]]"
   ]
  }
 ],
 "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
}