Document-Clustering-and-Topic-Modeling.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Document Clustering and Topic Modeling" 
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In this project, we use unsupervised learning models to cluster unlabeled documents into different groups, visualize the results and identify their latent topics/structures."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Contents"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<ul>\n",
    "<li>[Part 1: Load Data](#Part-1:-Load-Data)\n",
    "<li>[Part 2: Tokenizing and Stemming](#Part-2:-Tokenizing-and-Stemming)\n",
    "<li>[Part 3: TF-IDF](#Part-3:-TF-IDF)\n",
    "<li>[Part 4: K-means clustering](#Part-4:-K-means-clustering)\n",
    "<li>[Part 5: Topic Modeling - Latent Dirichlet Allocation](#Part-5:-Topic-Modeling---Latent-Dirichlet-Allocation)\n",
    "</ul>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Part 1: Load Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "import warnings\n",
    "warnings.filterwarnings('ignore')\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import nltk\n",
    "import re\n",
    "import os\n",
    "\n",
    "from sklearn import decomposition\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "import matplotlib.pyplot as plt\n",
    "import lda"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Read data from files. In summary, we have 100 titles and 100 synoposes (combined from imdb and wiki)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "#import three lists: titles and wikipedia synopses\n",
    "titles = open('../data/title_list.txt').read().split('\\n')\n",
    "titles = titles[:100] #ensures that only the first 100 are read in\n",
    "\n",
    "#The wiki synopses and imdb synopses of each movie is seperated by the keywords \"BREAKS HERE\". \n",
    "#Each synoposes may consist of multiple paragraphs.\n",
    "synopses_wiki = open('../data/synopses_list_wiki.txt').read().split('\\n BREAKS HERE')\n",
    "synopses_wiki = synopses_wiki[:100]\n",
    "\n",
    "synopses_imdb = open('../data/synopses_list_imdb.txt').read().split('\\n BREAKS HERE')\n",
    "synopses_imdb = synopses_imdb[:100]\n",
    "\n",
    "#Combine imdb and wiki to get full synoposes for the top 100 movies. \n",
    "synopses = []\n",
    "for i in range(len(synopses_wiki)):\n",
    "    item = synopses_wiki[i] + synopses_imdb[i]\n",
    "    synopses.append(item)\n",
    "    \n",
    "#Because these synopses have already been ordered in popularity order, \n",
    "#we just need to generate a list of ordered numbers for future usage.\n",
    "ranks = range(len(titles))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Part 2: Tokenizing and Stemming"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "Load stopwords and stemmer function from NLTK library.\n",
    "Stop words are words like \"a\", \"the\", or \"in\" which don't convey significant meaning.\n",
    "Stemming is the process of breaking a word down into its root."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "We use 179 stop-words from nltk library.\n",
      "[u'i', u'me', u'my', u'myself', u'we', u'our', u'ours', u'ourselves', u'you', u\"you're\"]\n"
     ]
    }
   ],
   "source": [
    "# Use nltk's English stopwords.\n",
    "stopwords = nltk.corpus.stopwords.words('english')\n",
    "\n",
    "print \"We use \" + str(len(stopwords)) + \" stop-words from nltk library.\"\n",
    "print stopwords[:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "from nltk.stem.snowball import SnowballStemmer\n",
    "stemmer = SnowballStemmer(\"english\")\n",
    "\n",
    "def tokenization_and_stemming(text):\n",
    "    tokens = [word.lower() for sent in nltk.sent_tokenize(text) for word in nltk.word_tokenize(sent) if word not in stopwords]\n",
    "#     tokens=[]\n",
    "#     for sent in nltk.sent_tokenize(text):\n",
    "#         for word in nltk.word_tokenize(sent):\n",
    "#             if word not in stopwords:\n",
    "#                 tokens.append(word);   \n",
    "    filtered_tokens = []\n",
    "    # filter out any tokens not containing letters (e.g., numeric tokens, raw punctuation)\n",
    "    for token in tokens:\n",
    "        if re.search('[a-zA-Z]', token):\n",
    "            filtered_tokens.append(token)\n",
    "    stems = [stemmer.stem(t) for t in filtered_tokens]\n",
    "    return stems\n",
    "\n",
    "def tokenization(text):\n",
    "    tokens = [word.lower() for sent in nltk.sent_tokenize(text) for word in nltk.word_tokenize(sent) if word not in stopwords]\n",
    "    filtered_tokens = []\n",
    "    for token in tokens:\n",
    "        if re.search('[a-zA-Z]', token):\n",
    "            filtered_tokens.append(token)\n",
    "    return filtered_tokens"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[u'look', u'father', \"'s\", u'arm']"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tokenization_and_stemming(\"she looked at her father's arm.\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Use our defined functions to analyze (i.e. tokenize, stem) our synoposes."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "docs_stemmed = []\n",
    "docs_tokenized = []\n",
    "for i in synopses:\n",
    "    tokenized_and_stemmed_results = tokenization_and_stemming(i)\n",
    "    docs_stemmed.extend(tokenized_and_stemmed_results)\n",
    "    \n",
    "    tokenized_results = tokenization(i)\n",
    "    docs_tokenized.extend(tokenized_results)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Create a mapping from stemmed words to original tokenized words for result interpretation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "angeles\n"
     ]
    }
   ],
   "source": [
    "vocab_frame_dict = {docs_stemmed[x]:docs_tokenized[x] for x in range(len(docs_stemmed))}\n",
    "print vocab_frame_dict['angel']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "# Part 3: TF-IDF"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In total, there are 100 synoposes and 538 terms.\n"
     ]
    }
   ],
   "source": [
    "#define vectorizer parameters\n",
    "tfidf_model = TfidfVectorizer(max_df=0.8, max_features=200000,\n",
    "                                 min_df=0.2, stop_words='english',\n",
    "                                 use_idf=True, tokenizer=tokenization_and_stemming, ngram_range=(1,1))\n",
    "\n",
    "tfidf_matrix = tfidf_model.fit_transform(synopses) #fit the vectorizer to synopses\n",
    "\n",
    "print \"In total, there are \" + str(tfidf_matrix.shape[0]) + \\\n",
    "      \" synoposes and \" + str(tfidf_matrix.shape[1]) + \" terms.\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'analyzer': u'word',\n",
       " 'binary': False,\n",
       " 'decode_error': u'strict',\n",
       " 'dtype': numpy.int64,\n",
       " 'encoding': u'utf-8',\n",
       " 'input': u'content',\n",
       " 'lowercase': True,\n",
       " 'max_df': 0.8,\n",
       " 'max_features': 200000,\n",
       " 'min_df': 0.2,\n",
       " 'ngram_range': (1, 1),\n",
       " 'norm': u'l2',\n",
       " 'preprocessor': None,\n",
       " 'smooth_idf': True,\n",
       " 'stop_words': 'english',\n",
       " 'strip_accents': None,\n",
       " 'sublinear_tf': False,\n",
       " 'token_pattern': u'(?u)\\\\b\\\\w\\\\w+\\\\b',\n",
       " 'tokenizer': <function __main__.tokenization_and_stemming>,\n",
       " 'use_idf': True,\n",
       " 'vocabulary': None}"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tfidf_model.get_params()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Save the terms identified by TF-IDF."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<100x538 sparse matrix of type '<type 'numpy.float64'>'\n",
       "\twith 19003 stored elements in Compressed Sparse Row format>"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tf_selected_words = tfidf_model.get_feature_names()\n",
    "tfidf_matrix"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Calculate Document Similarity"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[1.         0.1996283  0.23342595 ... 0.3581728  0.3277855  0.17714484]\n",
      " [0.1996283  1.         0.26626879 ... 0.26910321 0.22856581 0.10870124]\n",
      " [0.23342595 0.26626879 1.         ... 0.27738654 0.24669308 0.12805055]\n",
      " ...\n",
      " [0.3581728  0.26910321 0.27738654 ... 1.         0.55040973 0.08151334]\n",
      " [0.3277855  0.22856581 0.24669308 ... 0.55040973 1.         0.13005234]\n",
      " [0.17714484 0.10870124 0.12805055 ... 0.08151334 0.13005234 1.        ]]\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics.pairwise import cosine_similarity\n",
    "cos_matrix = cosine_similarity(tfidf_matrix)\n",
    "print cos_matrix"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Part 4: K-means clustering"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "from sklearn.cluster import KMeans\n",
    "\n",
    "num_clusters = 5\n",
    "km = KMeans(n_clusters=num_clusters)\n",
    "km.fit(tfidf_matrix)\n",
    "clusters = km.labels_.tolist()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4.1. Analyze K-means Result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create DataFrame films from all of the input files.\n",
    "films = { 'title': titles, 'rank': ranks, 'synopsis': synopses, 'cluster': clusters}\n",
    "frame = pd.DataFrame(films, index = [clusters] , columns = ['rank', 'title', 'cluster'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>rank</th>\n",
       "      <th>title</th>\n",
       "      <th>cluster</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>The Godfather</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>The Shawshank Redemption</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2</td>\n",
       "      <td>Schindler's List</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>Raging Bull</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>4</td>\n",
       "      <td>Casablanca</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>5</td>\n",
       "      <td>One Flew Over the Cuckoo's Nest</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>6</td>\n",
       "      <td>Gone with the Wind</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>7</td>\n",
       "      <td>Citizen Kane</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>8</td>\n",
       "      <td>The Wizard of Oz</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>9</td>\n",
       "      <td>Titanic</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   rank                            title  cluster\n",
       "2     0                    The Godfather        2\n",
       "1     1         The Shawshank Redemption        1\n",
       "3     2                 Schindler's List        3\n",
       "2     3                      Raging Bull        2\n",
       "1     4                       Casablanca        1\n",
       "3     5  One Flew Over the Cuckoo's Nest        3\n",
       "2     6               Gone with the Wind        2\n",
       "2     7                     Citizen Kane        2\n",
       "3     8                 The Wizard of Oz        3\n",
       "1     9                          Titanic        1"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "frame.head(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of films included in each cluster:\n"
     ]
    },
    {
     "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>cluster</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>26</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>25</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   cluster\n",
       "2       37\n",
       "1       26\n",
       "3       25\n",
       "4        6\n",
       "0        6"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print \"Number of films included in each cluster:\"\n",
    "frame['cluster'].value_counts().to_frame()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average rank (1 to 100) per cluster:\n"
     ]
    },
    {
     "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>rank</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>cluster</th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>54.166667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>53.884615</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>49.540541</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>42.960000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>52.833333</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              rank\n",
       "cluster           \n",
       "0        54.166667\n",
       "1        53.884615\n",
       "2        49.540541\n",
       "3        42.960000\n",
       "4        52.833333"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "grouped = frame['rank'].groupby(frame['cluster'])\n",
    "print \"Average rank (1 to 100) per cluster:\"\n",
    "grouped.mean().to_frame()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<Document clustering result by K-means>\n",
      "Cluster 0 words: filmed, singing, performed, love, house, problems,\n",
      "Cluster 0 titles (6 movies): \n",
      "Sunset Blvd., Singin' in the Rain, Gandhi, Good Will Hunting, Nashville, Yankee Doodle Dandy\n",
      "\n",
      "Cluster 1 words: police, car, killing, asks, say, murder,\n",
      "Cluster 1 titles (26 movies): \n",
      "The Shawshank Redemption, Casablanca, Titanic, Psycho, Vertigo, West Side Story, E.T. the Extra-Terrestrial, 2001: A Space Odyssey, The Silence of the Lambs, Chinatown, Some Like It Hot, Rocky, Jaws, The French Connection, Fargo, Shane, American Graffiti, Pulp Fiction, The Maltese Falcon, A Clockwork Orange, Taxi Driver, Double Indemnity, Rebel Without a Cause, Rear Window, The Third Man, North by Northwest\n",
      "\n",
      "Cluster 2 words: father, family, new, home, love, friend,\n",
      "Cluster 2 titles (37 movies): \n",
      "The Godfather, Raging Bull, Gone with the Wind, Citizen Kane, The Godfather: Part II, On the Waterfront, Forrest Gump, The Sound of Music, 12 Angry Men, Amadeus, A Streetcar Named Desire, To Kill a Mockingbird, The Best Years of Our Lives, My Fair Lady, Ben-Hur, Doctor Zhivago, Braveheart, The Good, the Bad and the Ugly, The Apartment, High Noon, The Pianist, Goodfellas, The Exorcist, City Lights, It Happened One Night, Midnight Cowboy, Mr. Smith Goes to Washington, Rain Man, Annie Hall, Out of Africa, Terms of Endearment, Giant, The Grapes of Wrath, The Green Mile, Network, The Graduate, Wuthering Heights\n",
      "\n",
      "Cluster 3 words: killing, soldiers, army, commanded, orders, men,\n",
      "Cluster 3 titles (25 movies): \n",
      "Schindler's List, One Flew Over the Cuckoo's Nest, The Wizard of Oz, Lawrence of Arabia, Star Wars, The Bridge on the River Kwai, Dr. Strangelove or: How I Learned to Stop Worrying and Love the Bomb, Apocalypse Now, The Lord of the Rings: The Return of the King, Gladiator, From Here to Eternity, Saving Private Ryan, Unforgiven, Raiders of the Lost Ark, Patton, Butch Cassidy and the Sundance Kid, The Treasure of the Sierra Madre, Platoon, Dances with Wolves, The Deer Hunter, All Quiet on the Western Front, Close Encounters of the Third Kind, The African Queen, Stagecoach, Mutiny on the Bounty\n",
      "\n",
      "Cluster 4 words: george, marries, family, woman, friend, john,\n",
      "Cluster 4 titles (6 movies): \n",
      "It's a Wonderful Life, The Philadelphia Story, An American in Paris, The King's Speech, A Place in the Sun, Tootsie\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print \"<Document clustering result by K-means>\"\n",
    "\n",
    "#km.cluster_centers_ denotes the importances of each items in centroid.\n",
    "#We need to sort it in decreasing-order and get the top k items.\n",
    "order_centroids = km.cluster_centers_.argsort()[:, ::-1] \n",
    "\n",
    "Cluster_keywords_summary = {}\n",
    "for i in range(num_clusters):\n",
    "    print \"Cluster \" + str(i) + \" words:\" ,\n",
    "    Cluster_keywords_summary[i] = []\n",
    "    for ind in order_centroids[i, :6]: #replace 6 with n words per cluster\n",
    "        Cluster_keywords_summary[i].append(vocab_frame_dict[tf_selected_words[ind]])\n",
    "        print vocab_frame_dict[tf_selected_words[ind]] + \",\",\n",
    "    print\n",
    "    #Here ix means index, which is the clusterID of each item.\n",
    "    #Without tolist, the values result from dataframe is <type 'numpy.ndarray'>\n",
    "    cluster_movies = frame.ix[i]['title'].values.tolist()\n",
    "    print \"Cluster \" + str(i) + \" titles (\" + str(len(cluster_movies)) + \" movies): \" \n",
    "    print \", \".join(cluster_movies)\n",
    "    print"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4.2. Plot K-means Result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "pca = decomposition.PCA(n_components=2)\n",
    "tfidf_matrix_np=tfidf_matrix.toarray()\n",
    "pca.fit(tfidf_matrix_np)\n",
    "X = pca.transform(tfidf_matrix_np)\n",
    "\n",
    "xs, ys = X[:, 0], X[:, 1]\n",
    "\n",
    "#set up colors per clusters using a dict\n",
    "cluster_colors = {0: '#1b9e77', 1: '#d95f02', 2: '#7570b3', 3: '#e7298a', 4: '#66a61e'}\n",
    "#set up cluster names using a dict\n",
    "cluster_names = {}\n",
    "for i in range(num_clusters):\n",
    "    cluster_names[i] = \", \".join(Cluster_keywords_summary[i])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x648 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline \n",
    "\n",
    "#create data frame with PCA cluster results\n",
    "df = pd.DataFrame(dict(x=xs, y=ys, label=clusters, title=titles)) \n",
    "groups = df.groupby(clusters)\n",
    "\n",
    "# set up plot\n",
    "fig, ax = plt.subplots(figsize=(16, 9))\n",
    "#Set color for each cluster/group\n",
    "for name, group in groups:\n",
    "    ax.plot(group.x, group.y, marker='o', linestyle='', ms=12, \n",
    "            label=cluster_names[name], color=cluster_colors[name], \n",
    "            mec='none')\n",
    "\n",
    "ax.legend(numpoints=1,loc=4)  #show legend with only 1 point, position is right bottom.\n",
    "\n",
    "plt.show() #show the plot"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.close()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Part 5: Topic Modeling - Latent Dirichlet Allocation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Now we use synopses to build a 100*551 matrix (terms)\n",
    "terms_dict = {tf_selected_words[x]:x for x in range(len(tf_selected_words))}\n",
    "\n",
    "feature_matrix_synopses_tf = []\n",
    "for i in synopses:\n",
    "    current_doc_stats = [0] * len(tf_selected_words)\n",
    "    allwords_stemmed = tokenization_and_stemming(i)\n",
    "    for get_terms in allwords_stemmed:\n",
    "        if get_terms in tf_selected_words:\n",
    "            current_doc_stats[terms_dict[get_terms]] += 1\n",
    "    current_doc_stats = np.asarray(current_doc_stats)\n",
    "    feature_matrix_synopses_tf.append(current_doc_stats)\n",
    "    \n",
    "feature_matrix_synopses_tf = np.asarray(feature_matrix_synopses_tf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = lda.LDA(n_topics=5, n_iter=500, random_state=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<li> \"model.topic_word_\" saves the importance of tf_selected_words in LDA model, i.e. words similarity matrix\n",
    "<li> The shape of \"model.topic_word_\" is (n_topics,num_of_selected_words)\n",
    "<li> \"model.doc_topic_\" saves the document topic results, i.e. document topic matrix.\n",
    "<li> The shape of \"model.doc_topic_\" is (num_of_documents, n_topics)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:lda:n_documents: 100\n",
      "INFO:lda:vocab_size: 538\n",
      "INFO:lda:n_words: 61218\n",
      "INFO:lda:n_topics: 5\n",
      "INFO:lda:n_iter: 500\n",
      "INFO:lda:<0> log likelihood: -484683\n",
      "INFO:lda:<10> log likelihood: -428330\n",
      "INFO:lda:<20> log likelihood: -412489\n",
      "INFO:lda:<30> log likelihood: -406790\n",
      "INFO:lda:<40> log likelihood: -403019\n",
      "INFO:lda:<50> log likelihood: -400070\n",
      "INFO:lda:<60> log likelihood: -398296\n",
      "INFO:lda:<70> log likelihood: -397266\n",
      "INFO:lda:<80> log likelihood: -395732\n",
      "INFO:lda:<90> log likelihood: -395476\n",
      "INFO:lda:<100> log likelihood: -394666\n",
      "INFO:lda:<110> log likelihood: -394617\n",
      "INFO:lda:<120> log likelihood: -393980\n",
      "INFO:lda:<130> log likelihood: -393121\n",
      "INFO:lda:<140> log likelihood: -392773\n",
      "INFO:lda:<150> log likelihood: -392559\n",
      "INFO:lda:<160> log likelihood: -392274\n",
      "INFO:lda:<170> log likelihood: -391875\n",
      "INFO:lda:<180> log likelihood: -392254\n",
      "INFO:lda:<190> log likelihood: -391966\n",
      "INFO:lda:<200> log likelihood: -391777\n",
      "INFO:lda:<210> log likelihood: -391467\n",
      "INFO:lda:<220> log likelihood: -391043\n",
      "INFO:lda:<230> log likelihood: -391624\n",
      "INFO:lda:<240> log likelihood: -391707\n",
      "INFO:lda:<250> log likelihood: -391696\n",
      "INFO:lda:<260> log likelihood: -391673\n",
      "INFO:lda:<270> log likelihood: -391540\n",
      "INFO:lda:<280> log likelihood: -391234\n",
      "INFO:lda:<290> log likelihood: -390801\n",
      "INFO:lda:<300> log likelihood: -391805\n",
      "INFO:lda:<310> log likelihood: -392021\n",
      "INFO:lda:<320> log likelihood: -391969\n",
      "INFO:lda:<330> log likelihood: -390905\n",
      "INFO:lda:<340> log likelihood: -391478\n",
      "INFO:lda:<350> log likelihood: -390938\n",
      "INFO:lda:<360> log likelihood: -390935\n",
      "INFO:lda:<370> log likelihood: -390741\n",
      "INFO:lda:<380> log likelihood: -391222\n",
      "INFO:lda:<390> log likelihood: -390952\n",
      "INFO:lda:<400> log likelihood: -390668\n",
      "INFO:lda:<410> log likelihood: -390701\n",
      "INFO:lda:<420> log likelihood: -390636\n",
      "INFO:lda:<430> log likelihood: -390205\n",
      "INFO:lda:<440> log likelihood: -390684\n",
      "INFO:lda:<450> log likelihood: -390519\n",
      "INFO:lda:<460> log likelihood: -390995\n",
      "INFO:lda:<470> log likelihood: -390638\n",
      "INFO:lda:<480> log likelihood: -390798\n",
      "INFO:lda:<490> log likelihood: -390844\n",
      "INFO:lda:<499> log likelihood: -390644\n"
     ]
    }
   ],
   "source": [
    "model.fit(feature_matrix_synopses_tf)\n",
    "topic_word = model.topic_word_  # model.components_ also works\n",
    "n_top_words = 7\n",
    "\n",
    "topic_keywords_list = []\n",
    "for i, topic_dist in enumerate(topic_word):\n",
    "    #Here we select top(n_top_words-1)\n",
    "    lda_topic_words = np.array(tf_selected_words)[np.argsort(topic_dist)][:-n_top_words:-1] \n",
    "    for j in range(len(lda_topic_words)):\n",
    "        lda_topic_words[j] = vocab_frame_dict[lda_topic_words[j]]\n",
    "    topic_keywords_list.append(lda_topic_words.tolist())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<Document clustering result by LDA>\n",
      "Cluster 0 words: george, war, soldiers, army, captain, commanded\n",
      "Cluster 0 titles (10 movies): \n",
      "Schindler's List, Lawrence of Arabia, It's a Wonderful Life, Dr. Strangelove or: How I Learned to Stop Worrying and Love the Bomb, Apocalypse Now, The Lord of the Rings: The Return of the King, Patton, Dances with Wolves, Stagecoach, Mutiny on the Bounty\n",
      "\n",
      "Cluster 1 words: car, away, police, meet, helping, apartment\n",
      "Cluster 1 titles (13 movies): \n",
      "Titanic, The Silence of the Lambs, The Apartment, The Pianist, The French Connection, It Happened One Night, Fargo, The Graduate, The Maltese Falcon, Double Indemnity, Rebel Without a Cause, Rear Window, North by Northwest\n",
      "\n",
      "Cluster 2 words: father, family, new, end, life, years\n",
      "Cluster 2 titles (28 movies): \n",
      "The Godfather, Casablanca, Citizen Kane, The Godfather: Part II, Sunset Blvd., Forrest Gump, The Sound of Music, Singin' in the Rain, Amadeus, A Streetcar Named Desire, The Best Years of Our Lives, Doctor Zhivago, Braveheart, The Good, the Bad and the Ugly, Goodfellas, City Lights, The King's Speech, A Place in the Sun, Midnight Cowboy, Mr. Smith Goes to Washington, Rain Man, Annie Hall, Out of Africa, Good Will Hunting, Terms of Endearment, Giant, Network, Yankee Doodle Dandy\n",
      "\n",
      "Cluster 3 words: n't, days, friend, comes, say, house\n",
      "Cluster 3 titles (26 movies): \n",
      "Raging Bull, One Flew Over the Cuckoo's Nest, Gone with the Wind, The Wizard of Oz, Psycho, Vertigo, West Side Story, E.T. the Extra-Terrestrial, Some Like It Hot, Gandhi, Rocky, The Philadelphia Story, An American in Paris, My Fair Lady, Ben-Hur, The Exorcist, All Quiet on the Western Front, Tootsie, The Grapes of Wrath, Nashville, American Graffiti, Pulp Fiction, The African Queen, A Clockwork Orange, Wuthering Heights, The Third Man\n",
      "\n",
      "Cluster 4 words: killing, death, men, use, john, shooting\n",
      "Cluster 4 titles (23 movies): \n",
      "The Shawshank Redemption, On the Waterfront, Star Wars, 2001: A Space Odyssey, Chinatown, The Bridge on the River Kwai, 12 Angry Men, Gladiator, From Here to Eternity, Saving Private Ryan, Unforgiven, Raiders of the Lost Ark, To Kill a Mockingbird, Jaws, Butch Cassidy and the Sundance Kid, The Treasure of the Sierra Madre, Platoon, High Noon, The Deer Hunter, Shane, The Green Mile, Close Encounters of the Third Kind, Taxi Driver\n",
      "\n"
     ]
    }
   ],
   "source": [
    "doc_topic = model.doc_topic_\n",
    "topic_doc_dict = {}\n",
    "\n",
    "print \"<Document clustering result by LDA>\"\n",
    "for i in range(len(doc_topic)):\n",
    "    topicID = doc_topic[i].argmax()\n",
    "    if topicID not in topic_doc_dict:\n",
    "        topic_doc_dict[topicID] = [titles[i]]\n",
    "    else:\n",
    "        topic_doc_dict[topicID].append(titles[i])\n",
    "for i in topic_doc_dict:\n",
    "    print \"Cluster \" + str(i) + \" words: \" + \", \".join(topic_keywords_list[i])\n",
    "    print \"Cluster \" + str(i) + \" titles (\" + str(len(topic_doc_dict[i])) + \" movies): \" \n",
    "    print ', '.join(topic_doc_dict[i])\n",
    "    print"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Homework: use LDA package from sklearn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.decomposition import LatentDirichletAllocation as LDA\n",
    "lda = LDA(n_components=5)  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5, 538)\n",
      "[[u'family', u'war', u'father', u'home', u'son', u'sister'], [u'love', u'apartment', u'dance', u'singing', u'performed', u'filmed'], [u'captain', u'office', u'ship', u'commanded', u'water', u'car'], [u'george', u'father', u'john', u'marries', u'house', u'brother'], [u'killing', u'soldiers', u'men', u'town', u'gun', u'train']]\n",
      "<Document clustering result by LDA>\n",
      "Cluster 0 words: family, war, father, home, son, sister\n",
      "Cluster 0 titles (10 movies): \n",
      "Schindler's List, Lawrence of Arabia, It's a Wonderful Life, Dr. Strangelove or: How I Learned to Stop Worrying and Love the Bomb, Apocalypse Now, The Lord of the Rings: The Return of the King, Patton, Dances with Wolves, Stagecoach, Mutiny on the Bounty\n",
      "\n",
      "Cluster 1 words: love, apartment, dance, singing, performed, filmed\n",
      "Cluster 1 titles (13 movies): \n",
      "Titanic, The Silence of the Lambs, The Apartment, The Pianist, The French Connection, It Happened One Night, Fargo, The Graduate, The Maltese Falcon, Double Indemnity, Rebel Without a Cause, Rear Window, North by Northwest\n",
      "\n",
      "Cluster 2 words: captain, office, ship, commanded, water, car\n",
      "Cluster 2 titles (28 movies): \n",
      "The Godfather, Casablanca, Citizen Kane, The Godfather: Part II, Sunset Blvd., Forrest Gump, The Sound of Music, Singin' in the Rain, Amadeus, A Streetcar Named Desire, The Best Years of Our Lives, Doctor Zhivago, Braveheart, The Good, the Bad and the Ugly, Goodfellas, City Lights, The King's Speech, A Place in the Sun, Midnight Cowboy, Mr. Smith Goes to Washington, Rain Man, Annie Hall, Out of Africa, Good Will Hunting, Terms of Endearment, Giant, Network, Yankee Doodle Dandy\n",
      "\n",
      "Cluster 3 words: george, father, john, marries, house, brother\n",
      "Cluster 3 titles (26 movies): \n",
      "Raging Bull, One Flew Over the Cuckoo's Nest, Gone with the Wind, The Wizard of Oz, Psycho, Vertigo, West Side Story, E.T. the Extra-Terrestrial, Some Like It Hot, Gandhi, Rocky, The Philadelphia Story, An American in Paris, My Fair Lady, Ben-Hur, The Exorcist, All Quiet on the Western Front, Tootsie, The Grapes of Wrath, Nashville, American Graffiti, Pulp Fiction, The African Queen, A Clockwork Orange, Wuthering Heights, The Third Man\n",
      "\n",
      "Cluster 4 words: killing, soldiers, men, town, gun, train\n",
      "Cluster 4 titles (23 movies): \n",
      "The Shawshank Redemption, On the Waterfront, Star Wars, 2001: A Space Odyssey, Chinatown, The Bridge on the River Kwai, 12 Angry Men, Gladiator, From Here to Eternity, Saving Private Ryan, Unforgiven, Raiders of the Lost Ark, To Kill a Mockingbird, Jaws, Butch Cassidy and the Sundance Kid, The Treasure of the Sierra Madre, Platoon, High Noon, The Deer Hunter, Shane, The Green Mile, Close Encounters of the Third Kind, Taxi Driver\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from sklearn.feature_extraction.text import TfidfVectorizer, CountVectorizer\n",
    "#this package won't perform well with feature vector of float type\n",
    "tfidf_matrix = tfidf_matrix.astype(int)\n",
    "\n",
    "lda.fit(tfidf_matrix)\n",
    "\n",
    "topic_word = lda.components_\n",
    "print topic_word.shape\n",
    "\n",
    "topic_keywords_list = []\n",
    "\n",
    "no_top_words = 10\n",
    "\n",
    "for i, topic_dist in enumerate(topic_word):\n",
    "    #Here we select top(n_top_words-1)\n",
    "    lda_topic_words = np.array(tf_selected_words)[np.argsort(topic_dist)][:-n_top_words:-1] \n",
    "    for j in range(len(lda_topic_words)):\n",
    "        lda_topic_words[j] = vocab_frame_dict[lda_topic_words[j]]\n",
    "    topic_keywords_list.append(lda_topic_words.tolist())\n",
    "print topic_keywords_list\n",
    "\n",
    "doc_topic = model.doc_topic_\n",
    "topic_doc_dict = {}\n",
    "\n",
    "print \"<Document clustering result by LDA>\"\n",
    "for i in range(len(doc_topic)):\n",
    "    topicID = doc_topic[i].argmax()\n",
    "    if topicID not in topic_doc_dict:\n",
    "        topic_doc_dict[topicID] = [titles[i]]\n",
    "    else:\n",
    "        topic_doc_dict[topicID].append(titles[i])\n",
    "for i in topic_doc_dict:\n",
    "    print \"Cluster \" + str(i) + \" words: \" + \", \".join(topic_keywords_list[i])\n",
    "    print \"Cluster \" + str(i) + \" titles (\" + str(len(topic_doc_dict[i])) + \" movies): \" \n",
    "    print ', '.join(topic_doc_dict[i])\n",
    "    print"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "# Appendix: K-means"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.datasets.samples_generator import make_blobs\n",
    "X, y = make_blobs(n_samples=300, centers=4,\n",
    "                  random_state=0, cluster_std=0.60)\n",
    "plt.scatter(X[:, 0], X[:, 1], s=50);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.cluster import KMeans\n",
    "est = KMeans(4)  # 4 clusters\n",
    "est.fit(X)\n",
    "y_kmeans = est.predict(X)\n",
    "plt.scatter(X[:, 0], X[:, 1], c=y_kmeans, s=50);"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.15"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}