{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Test classifier with all test dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_counts = vectorizer.transform(test_corpus)\n",
    "test_tfidf = transformer.transform(test_counts)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.944"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "classifier.score(test_tfidf, test_labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Find Informative Features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "n = 20\n",
    "feature_names = vectorizer.get_feature_names()  # Array mapping from feature integer indices to feature name"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AaronPressman 's Top 20 features \n",
      "\n",
      "products -6.843221656677828\n",
      "commerce -6.810211392169265\n",
      "phone -6.810211392169265\n",
      "administration -6.759908463882919\n",
      "securities -6.72589537844437\n",
      "federal -6.669434280670851\n",
      "court -6.656228165206716\n",
      "computer -6.648670161880032\n",
      "credit -6.59269344152553\n",
      "new -6.466038750420433\n",
      "companies -6.421187855927672\n",
      "banks -6.282188052280851\n",
      "congress -6.1691369178693325\n",
      "'s -5.874179559976962\n",
      "internet -5.48040317700573\n",
      "said -5.33729863768842\n",
      "'' -5.271277381900798\n",
      "`` -5.260227520807157\n",
      ". -4.229822040487652\n",
      ", -4.030488993297016\n",
      "\n",
      "\n",
      "AlanCrosby 's Top 20 features \n",
      "\n",
      "billion -6.651115217637997\n",
      "elections -6.56600473481026\n",
      "analysts -6.555871417225831\n",
      "match -6.527893548285615\n",
      "coalition -6.512588487497869\n",
      "week -6.388919463309094\n",
      "points -6.371236393412705\n",
      "index -6.340860179321937\n",
      "senate -6.295387040925007\n",
      "round -6.170196671090114\n",
      "( -6.150627605586323\n",
      ") -6.150627605586323\n",
      "percent -6.077633643022997\n",
      "market -6.022847352516765\n",
      "`` -5.770879251421359\n",
      "'' -5.745310055668684\n",
      "said -5.60866477446716\n",
      "'s -5.504211244742248\n",
      ". -4.294466466066362\n",
      ", -3.931324802675596\n",
      "\n",
      "\n",
      "AlexanderSmith 's Top 20 features \n",
      "\n",
      "cable -6.5614732431634435\n",
      "year -6.529223847082791\n",
      "investment -6.478300823188887\n",
      "british -6.463011028791085\n",
      "; -6.443924532066207\n",
      "$ -6.437760325964971\n",
      ") -6.38363052460965\n",
      "company -6.38363052460965\n",
      "( -6.3761314468987695\n",
      "group -6.3761314468987695\n",
      "percent -6.318081205830778\n",
      "& -6.296556432207335\n",
      "amp -6.296556432207335\n",
      "million -6.283439795017349\n",
      "'s -5.456188187197981\n",
      "`` -5.351676730265064\n",
      "'' -5.338355645441447\n",
      "said -5.139780496405044\n",
      ". -4.199710987553583\n",
      ", -4.107476056200915\n",
      "\n",
      "\n",
      "BenjaminKangLim 's Top 20 features \n",
      "\n",
      "( -6.818371181790302\n",
      ") -6.818371181790302\n",
      "state -6.79976981870553\n",
      "links -6.764806000459541\n",
      "dan -6.724913852417953\n",
      "people -6.710306847324062\n",
      "years -6.6989899973268265\n",
      "sentence -6.6316354347067294\n",
      "... -6.582288783415159\n",
      "government -6.495999271657508\n",
      "court -6.265417549885052\n",
      "party -6.18776190871761\n",
      "'' -5.5625764634394805\n",
      "`` -5.551799132659087\n",
      "beijing -5.454730907587863\n",
      "china -5.256907777517354\n",
      "said -5.081614300033448\n",
      "'s -5.068306343512768\n",
      ". -4.181346344826704\n",
      ", -4.087015543226064\n",
      "\n",
      "\n",
      "BernardHickey 's Top 20 features \n",
      "\n",
      "australian -6.496200125113844\n",
      "analysts -6.4895887628902855\n",
      "bank -6.465451563584756\n",
      "australia -6.330408486415948\n",
      "( -6.294269155165459\n",
      ") -6.294269155165459\n",
      "corp -6.252254460517263\n",
      "million -5.955480959210123\n",
      "profit -5.933900175980197\n",
      "year -5.921578211989878\n",
      "murdoch -5.905439558869782\n",
      "percent -5.851842538640983\n",
      "news -5.846674401805445\n",
      "$ -5.483310978602539\n",
      "'s -5.355618305216927\n",
      "`` -5.330694412171886\n",
      "'' -5.315426649102102\n",
      "said -4.8155626174681165\n",
      ". -4.2475736088130445\n",
      ", -4.224915439286149\n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "n = 20\n",
    "for i in range(5):\n",
    "    for j in label_encoder.inverse_transform([i]):\n",
    "        print(j, \"'s Top %s features \\n\" %n)\n",
    "        topn = sorted(zip(classifier.coef_[i], feature_names))[-n: ]\n",
    "        for coef, feature in topn:\n",
    "            print(feature, coef)\n",
    "        print(\"\\n\")"
   ]
  }
 ],
 "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}