AF_CS539_CVXOPT.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# packages to store and manipulate data\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "# plotting packages\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "\n",
    "# model building package\n",
    "import sklearn\n",
    "\n",
    "# package to clean text\n",
    "import re"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Exploring Data ##"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "data source: https://data.world/crowdflower/sentiment-of-climate-change"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv('climate_tweets.csv')"
   ]
  },
  {
   "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>tweet</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Global warming report urges governments to act...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Fighting poverty and global warming in Africa ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Carbon offsets: How a Vatican forest failed to...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Carbon offsets: How a Vatican forest failed to...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>URUGUAY: Tools Needed for Those Most Vulnerabl...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                               tweet\n",
       "0  Global warming report urges governments to act...\n",
       "1  Fighting poverty and global warming in Africa ...\n",
       "2  Carbon offsets: How a Vatican forest failed to...\n",
       "3  Carbon offsets: How a Vatican forest failed to...\n",
       "4  URUGUAY: Tools Needed for Those Most Vulnerabl..."
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(5541,)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.tweet.unique().shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(6090, 1)"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# two string variables for comparison\n",
    "string1 = 'climate'\n",
    "string2 = 'climb'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "False"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "string1 == string2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "string1[:4] == string2[:4]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# normal function example\n",
    "def my_normal_function(x):\n",
    "    return x**2 + 10\n",
    "# lambda function example\n",
    "my_lambda_function = lambda x: x**2 + 10"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "773"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# make a new column to highlight retweets\n",
    "df['is_retweet'] = df['tweet'].apply(lambda x: x[:2]=='RT')\n",
    "df['is_retweet'].sum()  # number of retweets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "677"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    " # number of unique retweets\n",
    "df.loc[df['is_retweet']].tweet.unique().size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "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>tweet</th>\n",
       "      <th>counts</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>3131</th>\n",
       "      <td>No matter if you believe in global warming or ...</td>\n",
       "      <td>20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4555</th>\n",
       "      <td>Take Action @change: Help Protect Wildlife Hab...</td>\n",
       "      <td>14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4027</th>\n",
       "      <td>RT @newtgingrich: Historic snow storm in washi...</td>\n",
       "      <td>9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1765</th>\n",
       "      <td>Fight Climate Change From All Fronts: [link]</td>\n",
       "      <td>8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1626</th>\n",
       "      <td>Earth's polar ice sheets vulnerable to even mo...</td>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1941</th>\n",
       "      <td>Global Warming Dent's El Ni'o's Protective Shi...</td>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1799</th>\n",
       "      <td>Foes of California's global warming law pour m...</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1351</th>\n",
       "      <td>Coalition of the Tired of Waiting: Fighting Cl...</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4271</th>\n",
       "      <td>SCIENCE: Scientists explore the evolution of c...</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1040</th>\n",
       "      <td>Carbon offsets: How a Vatican forest failed to...</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                  tweet  counts\n",
       "3131  No matter if you believe in global warming or ...      20\n",
       "4555  Take Action @change: Help Protect Wildlife Hab...      14\n",
       "4027  RT @newtgingrich: Historic snow storm in washi...       9\n",
       "1765       Fight Climate Change From All Fronts: [link]       8\n",
       "1626  Earth's polar ice sheets vulnerable to even mo...       7\n",
       "1941  Global Warming Dent's El Ni'o's Protective Shi...       7\n",
       "1799  Foes of California's global warming law pour m...       6\n",
       "1351  Coalition of the Tired of Waiting: Fighting Cl...       6\n",
       "4271  SCIENCE: Scientists explore the evolution of c...       6\n",
       "1040  Carbon offsets: How a Vatican forest failed to...       6"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 10 most repeated tweets\n",
    "df.groupby(['tweet']).size().reset_index(name='counts')\\\n",
    "  .sort_values('counts', ascending=False).head(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# number of times each tweet appears\n",
    "counts = df.groupby(['tweet']).size()\\\n",
    "           .reset_index(name='counts')\\\n",
    "           .counts\n",
    "\n",
    "# define bins for histogram\n",
    "my_bins = np.arange(0,counts.max()+2, 1)-0.5\n",
    "\n",
    "# plot histogram of tweet counts\n",
    "plt.figure()\n",
    "plt.hist(counts, bins = my_bins)\n",
    "plt.xlabels = np.arange(1,counts.max()+1, 1)\n",
    "plt.xlabel('copies of each tweet')\n",
    "plt.ylabel('frequency')\n",
    "plt.yscale('log', nonposy='clip')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "\"\"\"\n",
    "who is being retweeted (if any)\n",
    "who is being tweeted at/mentioned (if any)\n",
    "what hashtags are being used (if any)\n",
    "\"\"\"\n",
    "\n",
    "def find_retweeted(tweet):\n",
    "    '''This function will extract the twitter handles of retweed people'''\n",
    "    return re.findall('(?<=RT\\s)(@[A-Za-z]+[A-Za-z0-9-_]+)', tweet)\n",
    "\n",
    "def find_mentioned(tweet):\n",
    "    '''This function will extract the twitter handles of people mentioned in the tweet'''\n",
    "    return re.findall('(?<!RT\\s)(@[A-Za-z]+[A-Za-z0-9-_]+)', tweet)  \n",
    "\n",
    "def find_hashtags(tweet):\n",
    "    '''This function will extract hashtags'''\n",
    "    return re.findall('(#[A-Za-z]+[A-Za-z0-9-_]+)', tweet)  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# two sample tweets\n",
    "my_tweet = 'RT @our_codingclub: Can @you find #all the #hashtags?'\n",
    "my_other_tweet = 'Not a retweet. All views @my own'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['@our_codingclub']"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "find_retweeted(my_tweet)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['@you']"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "find_mentioned(my_tweet)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['#all', '#hashtags']"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "find_hashtags(my_tweet)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "# make new columns for retweeted usernames, mentioned usernames and hashtags\n",
    "df['retweeted'] = df.tweet.apply(find_retweeted)\n",
    "df['mentioned'] = df.tweet.apply(find_mentioned)\n",
    "df['hashtags'] = df.tweet.apply(find_hashtags)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "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>tweet</th>\n",
       "      <th>is_retweet</th>\n",
       "      <th>retweeted</th>\n",
       "      <th>mentioned</th>\n",
       "      <th>hashtags</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>6085</th>\n",
       "      <td>@bloodless_coup \"The phrase 'global warming' s...</td>\n",
       "      <td>False</td>\n",
       "      <td>[]</td>\n",
       "      <td>[@bloodless_coup]</td>\n",
       "      <td>[#p2, #tcot]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6086</th>\n",
       "      <td>Virginia to Investigate Global Warming Scienti...</td>\n",
       "      <td>False</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6087</th>\n",
       "      <td>Global warming you tube parody you will enjoy ...</td>\n",
       "      <td>False</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>[#IPCC, #ocra]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6088</th>\n",
       "      <td>One-Eyed Golfer: Don't dare tell me about glob...</td>\n",
       "      <td>False</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6089</th>\n",
       "      <td>man made global warming a hair brained theory ...</td>\n",
       "      <td>False</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>[#tcot, #p2, #climategate]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                  tweet  is_retweet retweeted  \\\n",
       "6085  @bloodless_coup \"The phrase 'global warming' s...       False        []   \n",
       "6086  Virginia to Investigate Global Warming Scienti...       False        []   \n",
       "6087  Global warming you tube parody you will enjoy ...       False        []   \n",
       "6088  One-Eyed Golfer: Don't dare tell me about glob...       False        []   \n",
       "6089  man made global warming a hair brained theory ...       False        []   \n",
       "\n",
       "              mentioned                    hashtags  \n",
       "6085  [@bloodless_coup]                [#p2, #tcot]  \n",
       "6086                 []                          []  \n",
       "6087                 []              [#IPCC, #ocra]  \n",
       "6088                 []                          []  \n",
       "6089                 []  [#tcot, #p2, #climategate]  "
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.tail()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "# take the rows from the hashtag columns where there are actually hashtags\n",
    "hashtags_list_df = df.loc[\n",
    "                       df.hashtags.apply(\n",
    "                           lambda hashtags_list: hashtags_list !=[]\n",
    "                       ),['hashtags']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "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>hashtags</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>[#Climate, #population]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>[#EarthDay]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>[#ac]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>[#tcot]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>36</th>\n",
       "      <td>[#tornadocot, #ocra, #sgp, #gop, #ucot, #tlot,...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                             hashtags\n",
       "12                            [#Climate, #population]\n",
       "16                                        [#EarthDay]\n",
       "26                                              [#ac]\n",
       "31                                            [#tcot]\n",
       "36  [#tornadocot, #ocra, #sgp, #gop, #ucot, #tlot,..."
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hashtags_list_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create dataframe where each use of hashtag gets its own row\n",
    "flattened_hashtags_df = pd.DataFrame(\n",
    "    [hashtag for hashtags_list in hashtags_list_df.hashtags\n",
    "    for hashtag in hashtags_list],\n",
    "    columns=['hashtag'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "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>hashtag</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>#Climate</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>#population</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>#EarthDay</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>#ac</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>#tcot</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       hashtag\n",
       "0     #Climate\n",
       "1  #population\n",
       "2    #EarthDay\n",
       "3          #ac\n",
       "4        #tcot"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "flattened_hashtags_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2067, 1)"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "flattened_hashtags_df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "477"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# number of unique hashtags\n",
    "flattened_hashtags_df['hashtag'].unique().size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "# count of appearances of each hashtag\n",
    "popular_hashtags = flattened_hashtags_df.groupby('hashtag').size()\\\n",
    "                                        .reset_index(name='counts')\\\n",
    "                                        .sort_values('counts', ascending=False)\\\n",
    "                                        .reset_index(drop=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "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>hashtag</th>\n",
       "      <th>counts</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>#tcot</td>\n",
       "      <td>227</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>#climate</td>\n",
       "      <td>202</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>#p2</td>\n",
       "      <td>112</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>#green</td>\n",
       "      <td>50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>#climatechange</td>\n",
       "      <td>47</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          hashtag  counts\n",
       "0           #tcot     227\n",
       "1        #climate     202\n",
       "2             #p2     112\n",
       "3          #green      50\n",
       "4  #climatechange      47"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "popular_hashtags.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# number of times each hashtag appears\n",
    "counts = flattened_hashtags_df.groupby(['hashtag']).size()\\\n",
    "                              .reset_index(name='counts')\\\n",
    "                              .counts\n",
    "\n",
    "# define bins for histogram                              \n",
    "my_bins = np.arange(0,counts.max()+2, 5)-0.5\n",
    "\n",
    "# plot histogram of tweet counts\n",
    "plt.figure()\n",
    "plt.hist(counts, bins = my_bins)\n",
    "plt.xlabels = np.arange(1,counts.max()+1, 1)\n",
    "plt.xlabel('hashtag number of appearances')\n",
    "plt.ylabel('frequency')\n",
    "plt.yscale('log', nonposy='clip')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## From Text to Vector ##"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "# take hashtags which appear at least this amount of times\n",
    "min_appearance = 10\n",
    "# find popular hashtags - make into python set for efficiency\n",
    "popular_hashtags_set = set(popular_hashtags[\n",
    "                           popular_hashtags.counts>=min_appearance\n",
    "                           ]['hashtag'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "# make a new column with only the popular hashtags\n",
    "hashtags_list_df['popular_hashtags'] = hashtags_list_df.hashtags.apply(\n",
    "            lambda hashtag_list: [hashtag for hashtag in hashtag_list\n",
    "                                  if hashtag in popular_hashtags_set])\n",
    "# drop rows without popular hashtag\n",
    "popular_hashtags_list_df = hashtags_list_df.loc[\n",
    "            hashtags_list_df.popular_hashtags.apply(lambda hashtag_list: hashtag_list !=[])]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "# make new dataframe\n",
    "hashtag_vector_df = popular_hashtags_list_df.loc[:, ['popular_hashtags']]\n",
    "\n",
    "for hashtag in popular_hashtags_set:\n",
    "    # make columns to encode presence of hashtags\n",
    "    hashtag_vector_df['{}'.format(hashtag)] = hashtag_vector_df.popular_hashtags.apply(\n",
    "        lambda hashtag_list: int(hashtag in hashtag_list))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "hashtag_matrix = hashtag_vector_df.drop('popular_hashtags', axis=1)\n",
    "\n",
    "# calculate the correlation matrix\n",
    "correlations = hashtag_matrix.corr()\n",
    "\n",
    "# plot the correlation matrix\n",
    "plt.figure(figsize=(10,10))\n",
    "sns.heatmap(correlations,\n",
    "    cmap='RdBu',\n",
    "    vmin=-1,\n",
    "    vmax=1,\n",
    "    square = True,\n",
    "    cbar_kws={'label':'correlation'})\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Topic Modelling ##"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "import nltk\n",
    "from nltk.tokenize import RegexpTokenizer\n",
    "from nltk.corpus import stopwords"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "# nltk.download('stopwords') # if it is your first time to use nltk, please unconmment and run it."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Cleaning Unstructured Text Data ##\n",
    "I would like to know the general things which people are talking about, not who they are talking about or to and not the web links they are sharing."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "def remove_links(tweet):\n",
    "    '''Takes a string and removes web links from it'''\n",
    "    tweet = re.sub(r'http\\S+', '', tweet) # remove http links\n",
    "    tweet = re.sub(r'bit.ly/\\S+', '', tweet) # rempve bitly links\n",
    "    tweet = tweet.strip('[link]') # remove [links]\n",
    "    return tweet\n",
    "\n",
    "def remove_users(tweet):\n",
    "    '''Takes a string and removes retweet and @user information'''\n",
    "    tweet = re.sub('(RT\\s@[A-Za-z]+[A-Za-z0-9-_]+)', '', tweet) # remove retweet\n",
    "    tweet = re.sub('(@[A-Za-z]+[A-Za-z0-9-_]+)', '', tweet) # remove tweeted at\n",
    "    return tweet"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "my_stopwords = nltk.corpus.stopwords.words('english')\n",
    "word_rooter = nltk.stem.snowball.PorterStemmer(ignore_stopwords=False).stem\n",
    "my_punctuation = '!\"$%&\\'()*+,-./:;<=>?[\\\\]^_`{|}~•@'\n",
    "\n",
    "# cleaning master function\n",
    "def clean_tweet(tweet, bigrams=False):\n",
    "    tweet = remove_users(tweet)\n",
    "    tweet = remove_links(tweet)\n",
    "    tweet = tweet.lower() # lower case\n",
    "    tweet = re.sub('['+my_punctuation + ']+', ' ', tweet) # strip punctuation\n",
    "    tweet = re.sub('\\s+', ' ', tweet) #remove double spacing\n",
    "    tweet = re.sub('([0-9]+)', '', tweet) # remove numbers\n",
    "    tweet_token_list = [word for word in tweet.split(' ')\n",
    "                            if word not in my_stopwords] # remove stopwords\n",
    "\n",
    "    tweet_token_list = [word_rooter(word) if '#' not in word else word\n",
    "                        for word in tweet_token_list] # apply word rooter\n",
    "    if bigrams:\n",
    "        tweet_token_list = tweet_token_list+[tweet_token_list[i]+'_'+tweet_token_list[i+1]\n",
    "                                            for i in range(len(tweet_token_list)-1)]\n",
    "    tweet = ' '.join(tweet_token_list)\n",
    "    return tweet"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "df['clean_tweet'] = df.tweet.apply(clean_tweet)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "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>tweet</th>\n",
       "      <th>is_retweet</th>\n",
       "      <th>retweeted</th>\n",
       "      <th>mentioned</th>\n",
       "      <th>hashtags</th>\n",
       "      <th>clean_tweet</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>6080</th>\n",
       "      <td>Bats, Birds and Lizards Can Fight Climate Chan...</td>\n",
       "      <td>False</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>bat bird lizard fight climat chang bird bat li...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6081</th>\n",
       "      <td>Bats, Birds and Lizards Can Fight Climate Chan...</td>\n",
       "      <td>False</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>bat bird lizard fight climat chang bird bat li...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6082</th>\n",
       "      <td>Global warming: The fossil fuel dilemma: Ameri...</td>\n",
       "      <td>False</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>global warm fossil fuel dilemma america use  m...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6083</th>\n",
       "      <td>RT @EPSclimate: Climate change row over the my...</td>\n",
       "      <td>True</td>\n",
       "      <td>[@EPSclimate]</td>\n",
       "      <td>[]</td>\n",
       "      <td>[#climate]</td>\n",
       "      <td>climat chang row mysteri shrink sheep #climate</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6084</th>\n",
       "      <td>It's 83•_Á and climbing in NYC. August weather...</td>\n",
       "      <td>False</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>á climb nyc august weather first day may unbe...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6085</th>\n",
       "      <td>@bloodless_coup \"The phrase 'global warming' s...</td>\n",
       "      <td>False</td>\n",
       "      <td>[]</td>\n",
       "      <td>[@bloodless_coup]</td>\n",
       "      <td>[#p2, #tcot]</td>\n",
       "      <td>phrase global warm abandon favor climat chang...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6086</th>\n",
       "      <td>Virginia to Investigate Global Warming Scienti...</td>\n",
       "      <td>False</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>virginia investig global warm scientist mann</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6087</th>\n",
       "      <td>Global warming you tube parody you will enjoy ...</td>\n",
       "      <td>False</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>[#IPCC, #ocra]</td>\n",
       "      <td>global warm tube parodi enjoy #ipcc #ocra</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6088</th>\n",
       "      <td>One-Eyed Golfer: Don't dare tell me about glob...</td>\n",
       "      <td>False</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>one eye golfer dare tell global warm twenti fi...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6089</th>\n",
       "      <td>man made global warming a hair brained theory ...</td>\n",
       "      <td>False</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>[#tcot, #p2, #climategate]</td>\n",
       "      <td>man made global warm hair brain theori  scient...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                  tweet  is_retweet  \\\n",
       "6080  Bats, Birds and Lizards Can Fight Climate Chan...       False   \n",
       "6081  Bats, Birds and Lizards Can Fight Climate Chan...       False   \n",
       "6082  Global warming: The fossil fuel dilemma: Ameri...       False   \n",
       "6083  RT @EPSclimate: Climate change row over the my...        True   \n",
       "6084  It's 83•_Á and climbing in NYC. August weather...       False   \n",
       "6085  @bloodless_coup \"The phrase 'global warming' s...       False   \n",
       "6086  Virginia to Investigate Global Warming Scienti...       False   \n",
       "6087  Global warming you tube parody you will enjoy ...       False   \n",
       "6088  One-Eyed Golfer: Don't dare tell me about glob...       False   \n",
       "6089  man made global warming a hair brained theory ...       False   \n",
       "\n",
       "          retweeted          mentioned                    hashtags  \\\n",
       "6080             []                 []                          []   \n",
       "6081             []                 []                          []   \n",
       "6082             []                 []                          []   \n",
       "6083  [@EPSclimate]                 []                  [#climate]   \n",
       "6084             []                 []                          []   \n",
       "6085             []  [@bloodless_coup]                [#p2, #tcot]   \n",
       "6086             []                 []                          []   \n",
       "6087             []                 []              [#IPCC, #ocra]   \n",
       "6088             []                 []                          []   \n",
       "6089             []                 []  [#tcot, #p2, #climategate]   \n",
       "\n",
       "                                            clean_tweet  \n",
       "6080  bat bird lizard fight climat chang bird bat li...  \n",
       "6081  bat bird lizard fight climat chang bird bat li...  \n",
       "6082  global warm fossil fuel dilemma america use  m...  \n",
       "6083     climat chang row mysteri shrink sheep #climate  \n",
       "6084   á climb nyc august weather first day may unbe...  \n",
       "6085   phrase global warm abandon favor climat chang...  \n",
       "6086      virginia investig global warm scientist mann   \n",
       "6087         global warm tube parodi enjoy #ipcc #ocra   \n",
       "6088  one eye golfer dare tell global warm twenti fi...  \n",
       "6089  man made global warm hair brain theori  scient...  "
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.tail(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[0.        0.        0.        ... 0.        0.        0.       ]\n",
      " [0.        0.        0.        ... 0.        0.        0.       ]\n",
      " [0.        0.        0.        ... 0.        0.        0.       ]\n",
      " ...\n",
      " [0.        0.        0.        ... 0.        0.        0.       ]\n",
      " [0.        0.        0.        ... 0.        0.        0.       ]\n",
      " [0.        0.        0.3549084 ... 0.        0.        0.       ]]\n"
     ]
    }
   ],
   "source": [
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "\n",
    "vectorizer = TfidfVectorizer(max_df=0.9, min_df=25, token_pattern='\\w+|\\$[\\d\\.]+|\\S+')\n",
    "\n",
    "# apply transformation\n",
    "tfidf = vectorizer.fit_transform(df['clean_tweet']).toarray()\n",
    "\n",
    "# tf_feature_names tells us what word each column in the matric represents\n",
    "tfidf_feature_names = vectorizer.get_feature_names() \n",
    "print(tfidf)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## LDA ##"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.decomposition import LatentDirichletAllocation\n",
    "\n",
    "number_of_topics = 10\n",
    "\n",
    "model_LDA = LatentDirichletAllocation(n_components=number_of_topics, random_state=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 5.95 s\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "LatentDirichletAllocation(batch_size=128, doc_topic_prior=None,\n",
       "                          evaluate_every=-1, learning_decay=0.7,\n",
       "                          learning_method='batch', learning_offset=10.0,\n",
       "                          max_doc_update_iter=100, max_iter=10,\n",
       "                          mean_change_tol=0.001, n_components=10, n_jobs=None,\n",
       "                          perp_tol=0.1, random_state=0, topic_word_prior=None,\n",
       "                          total_samples=1000000.0, verbose=0)"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%time model_LDA.fit(tfidf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "def display_topics(model, feature_names, no_top_words):\n",
    "    topic_dict = {}\n",
    "    for topic_idx, topic in enumerate(model.components_):\n",
    "        topic_dict[\"Topic %d words\" % (topic_idx+1)]= ['{}'.format(tfidf_feature_names[i])\n",
    "                        for i in topic.argsort()[:-no_top_words - 1:-1]]\n",
    "        topic_dict[\"Topic %d weights\" % (topic_idx+1)]= ['{:.3f}'.format(topic[i])\n",
    "                        for i in topic.argsort()[:-no_top_words - 1:-1]]\n",
    "    return pd.DataFrame(topic_dict)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "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>Topic 1 words</th>\n",
       "      <th>Topic 1 weights</th>\n",
       "      <th>Topic 2 words</th>\n",
       "      <th>Topic 2 weights</th>\n",
       "      <th>Topic 3 words</th>\n",
       "      <th>Topic 3 weights</th>\n",
       "      <th>Topic 4 words</th>\n",
       "      <th>Topic 4 weights</th>\n",
       "      <th>Topic 5 words</th>\n",
       "      <th>Topic 5 weights</th>\n",
       "      <th>Topic 6 words</th>\n",
       "      <th>Topic 6 weights</th>\n",
       "      <th>Topic 7 words</th>\n",
       "      <th>Topic 7 weights</th>\n",
       "      <th>Topic 8 words</th>\n",
       "      <th>Topic 8 weights</th>\n",
       "      <th>Topic 9 words</th>\n",
       "      <th>Topic 9 weights</th>\n",
       "      <th>Topic 10 words</th>\n",
       "      <th>Topic 10 weights</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>climat</td>\n",
       "      <td>134.120</td>\n",
       "      <td>global</td>\n",
       "      <td>87.362</td>\n",
       "      <td>global</td>\n",
       "      <td>171.479</td>\n",
       "      <td>global</td>\n",
       "      <td>64.383</td>\n",
       "      <td>climat</td>\n",
       "      <td>65.725</td>\n",
       "      <td>global</td>\n",
       "      <td>65.378</td>\n",
       "      <td>climat</td>\n",
       "      <td>70.921</td>\n",
       "      <td>chang</td>\n",
       "      <td>59.706</td>\n",
       "      <td>climat</td>\n",
       "      <td>51.415</td>\n",
       "      <td>climat</td>\n",
       "      <td>75.656</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>chang</td>\n",
       "      <td>130.114</td>\n",
       "      <td>warm</td>\n",
       "      <td>86.931</td>\n",
       "      <td>warm</td>\n",
       "      <td>169.887</td>\n",
       "      <td>warm</td>\n",
       "      <td>61.638</td>\n",
       "      <td>chang</td>\n",
       "      <td>62.225</td>\n",
       "      <td>warm</td>\n",
       "      <td>63.863</td>\n",
       "      <td>chang</td>\n",
       "      <td>68.960</td>\n",
       "      <td>climat</td>\n",
       "      <td>51.175</td>\n",
       "      <td>global</td>\n",
       "      <td>49.527</td>\n",
       "      <td>chang</td>\n",
       "      <td>74.613</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>via</td>\n",
       "      <td>34.821</td>\n",
       "      <td>snow</td>\n",
       "      <td>64.096</td>\n",
       "      <td>scientist</td>\n",
       "      <td>75.639</td>\n",
       "      <td>climat</td>\n",
       "      <td>33.723</td>\n",
       "      <td>warm</td>\n",
       "      <td>37.421</td>\n",
       "      <td>al</td>\n",
       "      <td>42.406</td>\n",
       "      <td>make</td>\n",
       "      <td>49.356</td>\n",
       "      <td>new</td>\n",
       "      <td>50.537</td>\n",
       "      <td>chang</td>\n",
       "      <td>48.570</td>\n",
       "      <td>confer</td>\n",
       "      <td>49.924</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>believ</td>\n",
       "      <td>31.404</td>\n",
       "      <td>like</td>\n",
       "      <td>47.689</td>\n",
       "      <td>climat</td>\n",
       "      <td>61.453</td>\n",
       "      <td>chang</td>\n",
       "      <td>32.435</td>\n",
       "      <td>via</td>\n",
       "      <td>36.989</td>\n",
       "      <td>day</td>\n",
       "      <td>42.037</td>\n",
       "      <td>energi</td>\n",
       "      <td>44.468</td>\n",
       "      <td>#climate</td>\n",
       "      <td>46.204</td>\n",
       "      <td>warm</td>\n",
       "      <td>46.465</td>\n",
       "      <td>peopl</td>\n",
       "      <td>49.898</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>global</td>\n",
       "      <td>30.757</td>\n",
       "      <td>volcano</td>\n",
       "      <td>36.359</td>\n",
       "      <td>scienc</td>\n",
       "      <td>56.110</td>\n",
       "      <td>come</td>\n",
       "      <td>30.083</td>\n",
       "      <td>global</td>\n",
       "      <td>36.780</td>\n",
       "      <td>gore</td>\n",
       "      <td>39.706</td>\n",
       "      <td>wors</td>\n",
       "      <td>37.890</td>\n",
       "      <td>agenc</td>\n",
       "      <td>39.543</td>\n",
       "      <td>great</td>\n",
       "      <td>44.869</td>\n",
       "      <td>talk</td>\n",
       "      <td>42.278</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>#tcot</td>\n",
       "      <td>29.071</td>\n",
       "      <td>dc</td>\n",
       "      <td>32.482</td>\n",
       "      <td>chang</td>\n",
       "      <td>49.362</td>\n",
       "      <td>know</td>\n",
       "      <td>29.599</td>\n",
       "      <td>news</td>\n",
       "      <td>34.838</td>\n",
       "      <td>one</td>\n",
       "      <td>35.897</td>\n",
       "      <td>allergi</td>\n",
       "      <td>31.913</td>\n",
       "      <td>take</td>\n",
       "      <td>39.003</td>\n",
       "      <td>senat</td>\n",
       "      <td>43.849</td>\n",
       "      <td>world</td>\n",
       "      <td>36.694</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>warm</td>\n",
       "      <td>29.039</td>\n",
       "      <td>storm</td>\n",
       "      <td>30.913</td>\n",
       "      <td>skeptic</td>\n",
       "      <td>32.083</td>\n",
       "      <td>green</td>\n",
       "      <td>27.644</td>\n",
       "      <td>good</td>\n",
       "      <td>31.417</td>\n",
       "      <td>effect</td>\n",
       "      <td>32.429</td>\n",
       "      <td>global</td>\n",
       "      <td>28.056</td>\n",
       "      <td>obama</td>\n",
       "      <td>35.820</td>\n",
       "      <td>bill</td>\n",
       "      <td>42.927</td>\n",
       "      <td>graham</td>\n",
       "      <td>34.482</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>trial</td>\n",
       "      <td>28.010</td>\n",
       "      <td>chang</td>\n",
       "      <td>30.280</td>\n",
       "      <td>un</td>\n",
       "      <td>31.791</td>\n",
       "      <td>tell</td>\n",
       "      <td>25.632</td>\n",
       "      <td>report</td>\n",
       "      <td>30.258</td>\n",
       "      <td>earth</td>\n",
       "      <td>31.801</td>\n",
       "      <td>warm</td>\n",
       "      <td>26.958</td>\n",
       "      <td>warm</td>\n",
       "      <td>33.099</td>\n",
       "      <td>stop</td>\n",
       "      <td>36.502</td>\n",
       "      <td>fight</td>\n",
       "      <td>32.875</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>clinic</td>\n",
       "      <td>28.010</td>\n",
       "      <td>climat</td>\n",
       "      <td>28.833</td>\n",
       "      <td>debat</td>\n",
       "      <td>28.315</td>\n",
       "      <td>caus</td>\n",
       "      <td>24.949</td>\n",
       "      <td>human</td>\n",
       "      <td>30.099</td>\n",
       "      <td>watch</td>\n",
       "      <td>25.111</td>\n",
       "      <td>expert</td>\n",
       "      <td>26.720</td>\n",
       "      <td>feder</td>\n",
       "      <td>30.161</td>\n",
       "      <td>immigr</td>\n",
       "      <td>31.104</td>\n",
       "      <td>bolivia</td>\n",
       "      <td>30.846</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>live</td>\n",
       "      <td>24.626</td>\n",
       "      <td>iceland</td>\n",
       "      <td>25.198</td>\n",
       "      <td>real</td>\n",
       "      <td>25.423</td>\n",
       "      <td>com</td>\n",
       "      <td>24.526</td>\n",
       "      <td>thing</td>\n",
       "      <td>27.797</td>\n",
       "      <td>chang</td>\n",
       "      <td>22.990</td>\n",
       "      <td>public</td>\n",
       "      <td>24.390</td>\n",
       "      <td>global</td>\n",
       "      <td>29.271</td>\n",
       "      <td>tackl</td>\n",
       "      <td>28.307</td>\n",
       "      <td>bill</td>\n",
       "      <td>28.074</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  Topic 1 words Topic 1 weights Topic 2 words Topic 2 weights Topic 3 words  \\\n",
       "0        climat         134.120        global          87.362        global   \n",
       "1         chang         130.114          warm          86.931          warm   \n",
       "2           via          34.821          snow          64.096     scientist   \n",
       "3        believ          31.404          like          47.689        climat   \n",
       "4        global          30.757       volcano          36.359        scienc   \n",
       "5         #tcot          29.071            dc          32.482         chang   \n",
       "6          warm          29.039         storm          30.913       skeptic   \n",
       "7         trial          28.010         chang          30.280            un   \n",
       "8        clinic          28.010        climat          28.833         debat   \n",
       "9          live          24.626       iceland          25.198          real   \n",
       "\n",
       "  Topic 3 weights Topic 4 words Topic 4 weights Topic 5 words Topic 5 weights  \\\n",
       "0         171.479        global          64.383        climat          65.725   \n",
       "1         169.887          warm          61.638         chang          62.225   \n",
       "2          75.639        climat          33.723          warm          37.421   \n",
       "3          61.453         chang          32.435           via          36.989   \n",
       "4          56.110          come          30.083        global          36.780   \n",
       "5          49.362          know          29.599          news          34.838   \n",
       "6          32.083         green          27.644          good          31.417   \n",
       "7          31.791          tell          25.632        report          30.258   \n",
       "8          28.315          caus          24.949         human          30.099   \n",
       "9          25.423           com          24.526         thing          27.797   \n",
       "\n",
       "  Topic 6 words Topic 6 weights Topic 7 words Topic 7 weights Topic 8 words  \\\n",
       "0        global          65.378        climat          70.921         chang   \n",
       "1          warm          63.863         chang          68.960        climat   \n",
       "2            al          42.406          make          49.356           new   \n",
       "3           day          42.037        energi          44.468      #climate   \n",
       "4          gore          39.706          wors          37.890         agenc   \n",
       "5           one          35.897       allergi          31.913          take   \n",
       "6        effect          32.429        global          28.056         obama   \n",
       "7         earth          31.801          warm          26.958          warm   \n",
       "8         watch          25.111        expert          26.720         feder   \n",
       "9         chang          22.990        public          24.390        global   \n",
       "\n",
       "  Topic 8 weights Topic 9 words Topic 9 weights Topic 10 words  \\\n",
       "0          59.706        climat          51.415         climat   \n",
       "1          51.175        global          49.527          chang   \n",
       "2          50.537         chang          48.570         confer   \n",
       "3          46.204          warm          46.465          peopl   \n",
       "4          39.543         great          44.869           talk   \n",
       "5          39.003         senat          43.849          world   \n",
       "6          35.820          bill          42.927         graham   \n",
       "7          33.099          stop          36.502          fight   \n",
       "8          30.161        immigr          31.104        bolivia   \n",
       "9          29.271         tackl          28.307           bill   \n",
       "\n",
       "  Topic 10 weights  \n",
       "0           75.656  \n",
       "1           74.613  \n",
       "2           49.924  \n",
       "3           49.898  \n",
       "4           42.278  \n",
       "5           36.694  \n",
       "6           34.482  \n",
       "7           32.875  \n",
       "8           30.846  \n",
       "9           28.074  "
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "no_top_words = 10\n",
    "display_topics(model_LDA, tfidf_feature_names, no_top_words)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "table_LDA = display_topics(model_LDA, tfidf_feature_names, no_top_words)\n",
    "table_LDA.to_csv(\"LDA.csv\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## NMF ##"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.decomposition import NMF\n",
    "\n",
    "number_of_topics = 10\n",
    "\n",
    "model_NMF = NMF(n_components=number_of_topics, random_state=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 251 ms\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "NMF(alpha=0.0, beta_loss='frobenius', init=None, l1_ratio=0.0, max_iter=200,\n",
       "    n_components=10, random_state=0, shuffle=False, solver='cd', tol=0.0001,\n",
       "    verbose=0)"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%time model_NMF.fit(tfidf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "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>Topic 1 words</th>\n",
       "      <th>Topic 1 weights</th>\n",
       "      <th>Topic 2 words</th>\n",
       "      <th>Topic 2 weights</th>\n",
       "      <th>Topic 3 words</th>\n",
       "      <th>Topic 3 weights</th>\n",
       "      <th>Topic 4 words</th>\n",
       "      <th>Topic 4 weights</th>\n",
       "      <th>Topic 5 words</th>\n",
       "      <th>Topic 5 weights</th>\n",
       "      <th>Topic 6 words</th>\n",
       "      <th>Topic 6 weights</th>\n",
       "      <th>Topic 7 words</th>\n",
       "      <th>Topic 7 weights</th>\n",
       "      <th>Topic 8 words</th>\n",
       "      <th>Topic 8 weights</th>\n",
       "      <th>Topic 9 words</th>\n",
       "      <th>Topic 9 weights</th>\n",
       "      <th>Topic 10 words</th>\n",
       "      <th>Topic 10 weights</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>global</td>\n",
       "      <td>3.664</td>\n",
       "      <td>climat</td>\n",
       "      <td>2.644</td>\n",
       "      <td>via</td>\n",
       "      <td>3.045</td>\n",
       "      <td>bill</td>\n",
       "      <td>2.815</td>\n",
       "      <td>new</td>\n",
       "      <td>2.137</td>\n",
       "      <td>snow</td>\n",
       "      <td>2.087</td>\n",
       "      <td>peopl</td>\n",
       "      <td>1.778</td>\n",
       "      <td>scienc</td>\n",
       "      <td>2.427</td>\n",
       "      <td>#climate</td>\n",
       "      <td>2.511</td>\n",
       "      <td>make</td>\n",
       "      <td>1.565</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>warm</td>\n",
       "      <td>3.614</td>\n",
       "      <td>chang</td>\n",
       "      <td>2.530</td>\n",
       "      <td>news</td>\n",
       "      <td>0.584</td>\n",
       "      <td>senat</td>\n",
       "      <td>1.895</td>\n",
       "      <td>agenc</td>\n",
       "      <td>0.851</td>\n",
       "      <td>gore</td>\n",
       "      <td>1.033</td>\n",
       "      <td>world</td>\n",
       "      <td>1.545</td>\n",
       "      <td>scientist</td>\n",
       "      <td>1.948</td>\n",
       "      <td>chang</td>\n",
       "      <td>0.615</td>\n",
       "      <td>wors</td>\n",
       "      <td>1.130</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>caus</td>\n",
       "      <td>0.218</td>\n",
       "      <td>fight</td>\n",
       "      <td>0.196</td>\n",
       "      <td>humanitarian</td>\n",
       "      <td>0.289</td>\n",
       "      <td>graham</td>\n",
       "      <td>1.064</td>\n",
       "      <td>feder</td>\n",
       "      <td>0.627</td>\n",
       "      <td>al</td>\n",
       "      <td>1.007</td>\n",
       "      <td>confer</td>\n",
       "      <td>1.327</td>\n",
       "      <td>debat</td>\n",
       "      <td>0.483</td>\n",
       "      <td>#global</td>\n",
       "      <td>0.469</td>\n",
       "      <td>blizzard</td>\n",
       "      <td>0.825</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>effect</td>\n",
       "      <td>0.174</td>\n",
       "      <td>energi</td>\n",
       "      <td>0.129</td>\n",
       "      <td>india</td>\n",
       "      <td>0.163</td>\n",
       "      <td>immigr</td>\n",
       "      <td>0.771</td>\n",
       "      <td>form</td>\n",
       "      <td>0.506</td>\n",
       "      <td>dc</td>\n",
       "      <td>0.883</td>\n",
       "      <td>earth</td>\n",
       "      <td>1.055</td>\n",
       "      <td>say</td>\n",
       "      <td>0.312</td>\n",
       "      <td>#eco</td>\n",
       "      <td>0.372</td>\n",
       "      <td>time</td>\n",
       "      <td>0.664</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>great</td>\n",
       "      <td>0.165</td>\n",
       "      <td>take</td>\n",
       "      <td>0.112</td>\n",
       "      <td>us</td>\n",
       "      <td>0.139</td>\n",
       "      <td>compromis</td>\n",
       "      <td>0.736</td>\n",
       "      <td>obama</td>\n",
       "      <td>0.425</td>\n",
       "      <td>#tcot</td>\n",
       "      <td>0.793</td>\n",
       "      <td>right</td>\n",
       "      <td>0.729</td>\n",
       "      <td>report</td>\n",
       "      <td>0.245</td>\n",
       "      <td>great</td>\n",
       "      <td>0.208</td>\n",
       "      <td>allergi</td>\n",
       "      <td>0.548</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>stop</td>\n",
       "      <td>0.152</td>\n",
       "      <td>legisl</td>\n",
       "      <td>0.111</td>\n",
       "      <td>polit</td>\n",
       "      <td>0.107</td>\n",
       "      <td>prepar</td>\n",
       "      <td>0.715</td>\n",
       "      <td>report</td>\n",
       "      <td>0.370</td>\n",
       "      <td>mean</td>\n",
       "      <td>0.530</td>\n",
       "      <td>bolivia</td>\n",
       "      <td>0.653</td>\n",
       "      <td>climat</td>\n",
       "      <td>0.192</td>\n",
       "      <td>#p</td>\n",
       "      <td>0.199</td>\n",
       "      <td>snowstorm</td>\n",
       "      <td>0.477</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>volcano</td>\n",
       "      <td>0.146</td>\n",
       "      <td>green</td>\n",
       "      <td>0.106</td>\n",
       "      <td>com</td>\n",
       "      <td>0.105</td>\n",
       "      <td>put</td>\n",
       "      <td>0.665</td>\n",
       "      <td>studi</td>\n",
       "      <td>0.340</td>\n",
       "      <td>storm</td>\n",
       "      <td>0.523</td>\n",
       "      <td>mother</td>\n",
       "      <td>0.576</td>\n",
       "      <td>palin</td>\n",
       "      <td>0.186</td>\n",
       "      <td>us</td>\n",
       "      <td>0.174</td>\n",
       "      <td>get</td>\n",
       "      <td>0.434</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>could</td>\n",
       "      <td>0.142</td>\n",
       "      <td>carbon</td>\n",
       "      <td>0.102</td>\n",
       "      <td>climat</td>\n",
       "      <td>0.097</td>\n",
       "      <td>limbo</td>\n",
       "      <td>0.518</td>\n",
       "      <td>york</td>\n",
       "      <td>0.292</td>\n",
       "      <td>washington</td>\n",
       "      <td>0.382</td>\n",
       "      <td>day</td>\n",
       "      <td>0.545</td>\n",
       "      <td>un</td>\n",
       "      <td>0.177</td>\n",
       "      <td>#green</td>\n",
       "      <td>0.155</td>\n",
       "      <td>c</td>\n",
       "      <td>0.417</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>believ</td>\n",
       "      <td>0.135</td>\n",
       "      <td>u</td>\n",
       "      <td>0.101</td>\n",
       "      <td>need</td>\n",
       "      <td>0.097</td>\n",
       "      <td>exit</td>\n",
       "      <td>0.498</td>\n",
       "      <td>administr</td>\n",
       "      <td>0.283</td>\n",
       "      <td>cold</td>\n",
       "      <td>0.376</td>\n",
       "      <td>indigen</td>\n",
       "      <td>0.306</td>\n",
       "      <td>snake</td>\n",
       "      <td>0.175</td>\n",
       "      <td>take</td>\n",
       "      <td>0.141</td>\n",
       "      <td>blame</td>\n",
       "      <td>0.272</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>like</td>\n",
       "      <td>0.134</td>\n",
       "      <td>nation</td>\n",
       "      <td>0.090</td>\n",
       "      <td>chang</td>\n",
       "      <td>0.095</td>\n",
       "      <td>legisl</td>\n",
       "      <td>0.473</td>\n",
       "      <td>propos</td>\n",
       "      <td>0.256</td>\n",
       "      <td>jr</td>\n",
       "      <td>0.328</td>\n",
       "      <td>summit</td>\n",
       "      <td>0.220</td>\n",
       "      <td>oil</td>\n",
       "      <td>0.171</td>\n",
       "      <td>fact</td>\n",
       "      <td>0.111</td>\n",
       "      <td>season</td>\n",
       "      <td>0.256</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  Topic 1 words Topic 1 weights Topic 2 words Topic 2 weights Topic 3 words  \\\n",
       "0        global           3.664        climat           2.644           via   \n",
       "1          warm           3.614         chang           2.530          news   \n",
       "2          caus           0.218         fight           0.196  humanitarian   \n",
       "3        effect           0.174        energi           0.129         india   \n",
       "4         great           0.165          take           0.112            us   \n",
       "5          stop           0.152        legisl           0.111         polit   \n",
       "6       volcano           0.146         green           0.106           com   \n",
       "7         could           0.142        carbon           0.102        climat   \n",
       "8        believ           0.135             u           0.101          need   \n",
       "9          like           0.134        nation           0.090         chang   \n",
       "\n",
       "  Topic 3 weights Topic 4 words Topic 4 weights Topic 5 words Topic 5 weights  \\\n",
       "0           3.045          bill           2.815           new           2.137   \n",
       "1           0.584         senat           1.895         agenc           0.851   \n",
       "2           0.289        graham           1.064         feder           0.627   \n",
       "3           0.163        immigr           0.771          form           0.506   \n",
       "4           0.139     compromis           0.736         obama           0.425   \n",
       "5           0.107        prepar           0.715        report           0.370   \n",
       "6           0.105           put           0.665         studi           0.340   \n",
       "7           0.097         limbo           0.518          york           0.292   \n",
       "8           0.097          exit           0.498     administr           0.283   \n",
       "9           0.095        legisl           0.473        propos           0.256   \n",
       "\n",
       "  Topic 6 words Topic 6 weights Topic 7 words Topic 7 weights Topic 8 words  \\\n",
       "0          snow           2.087         peopl           1.778        scienc   \n",
       "1          gore           1.033         world           1.545     scientist   \n",
       "2            al           1.007        confer           1.327         debat   \n",
       "3            dc           0.883         earth           1.055           say   \n",
       "4         #tcot           0.793         right           0.729        report   \n",
       "5          mean           0.530       bolivia           0.653        climat   \n",
       "6         storm           0.523        mother           0.576         palin   \n",
       "7    washington           0.382           day           0.545            un   \n",
       "8          cold           0.376       indigen           0.306         snake   \n",
       "9            jr           0.328        summit           0.220           oil   \n",
       "\n",
       "  Topic 8 weights Topic 9 words Topic 9 weights Topic 10 words  \\\n",
       "0           2.427      #climate           2.511           make   \n",
       "1           1.948         chang           0.615           wors   \n",
       "2           0.483       #global           0.469       blizzard   \n",
       "3           0.312          #eco           0.372           time   \n",
       "4           0.245         great           0.208        allergi   \n",
       "5           0.192            #p           0.199      snowstorm   \n",
       "6           0.186            us           0.174            get   \n",
       "7           0.177        #green           0.155              c   \n",
       "8           0.175          take           0.141          blame   \n",
       "9           0.171          fact           0.111         season   \n",
       "\n",
       "  Topic 10 weights  \n",
       "0            1.565  \n",
       "1            1.130  \n",
       "2            0.825  \n",
       "3            0.664  \n",
       "4            0.548  \n",
       "5            0.477  \n",
       "6            0.434  \n",
       "7            0.417  \n",
       "8            0.272  \n",
       "9            0.256  "
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "no_top_words = 10\n",
    "display_topics(model_NMF, tfidf_feature_names, no_top_words)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "table_NMF = display_topics(model_NMF, tfidf_feature_names, no_top_words)\n",
    "table_NMF.to_csv(\"NMF.csv\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Anchor Free Algorithm ##"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(425, 425)\n"
     ]
    }
   ],
   "source": [
    "from scipy.sparse import spdiags\n",
    "\n",
    "# ---- NCW procedure ------- #\n",
    "# References: Wei Xu, Xin Liu, and Yihong Gong.Document clustering Based on Non-negative Matrix Factorization.\n",
    "#             In Proceedings of the 26th annual international ACM SIGIR conference on Research and developmentin informaion retrieval, \n",
    "#             pages 267–273. ACM, 2003.\n",
    "    # remove zero rows\n",
    "    # np.where(~tfidf.any(axis=1))[0]\n",
    "    # a[~(a==0).all(1)]\n",
    "tfidf = tfidf[~(tfidf==0).all(1)]\n",
    "D_half = sum(tfidf*tfidf)**(-0.5)\n",
    "doc_num = D_half.shape[0] \n",
    "D_half = spdiags(D_half, 0, doc_num, doc_num)\n",
    "tfidf = tfidf * D_half\n",
    "\n",
    "# compute the correlation\n",
    "D = tfidf\n",
    "# print(D.shape)\n",
    "nWords, nDoc = D.shape\n",
    "P = (1/nDoc) * np.dot(D.T, D)\n",
    "print(P.shape)  # (425, 425)\n",
    "\n",
    "F = 10 # number of topics\n",
    "\n",
    "from scipy.sparse.linalg import eigs\n",
    "# Applying Eigen decomposition to D matrix\n",
    "\n",
    "vals, vecs = eigs(P, F)\n",
    "\n",
    "vals = np.real(vals)\n",
    "DD = np.diag(vals)\n",
    "# print(DD)\n",
    "DD = np.abs(DD)\n",
    "DD_root = np.sqrt(DD)\n",
    "B = np.real(np.dot(vecs, DD_root))\n",
    "# print(B.shape)\n",
    "# print(B)\n",
    "\n",
    "# Initialize Identity E\n",
    "E = np.identity(F)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████████████████████| 10/10 [00:02<00:00,  4.34it/s]\n",
      "100%|██████████████████████████████████████████████████████████████████████████████████| 10/10 [00:02<00:00,  4.33it/s]\n",
      "100%|██████████████████████████████████████████████████████████████████████████████████| 10/10 [00:02<00:00,  4.22it/s]\n"
     ]
    }
   ],
   "source": [
    "import cvxpy as cp\n",
    "# def AnchorFree(B, E):\n",
    "p = B.shape[0]\n",
    "n = B.shape[1]\n",
    "M = E\n",
    "# print(B)\n",
    "a = np.zeros(n)\n",
    "mm = 0\n",
    "conv = []\n",
    "\n",
    "from tqdm import tqdm\n",
    "\n",
    "while True:\n",
    "    for i in tqdm(range(n)):\n",
    "        for j in range(n):\n",
    "            step = (-1.)**(i+j)\n",
    "            N = np.delete(M, j, axis=0)\n",
    "            N = np.delete(N, i, axis=1)\n",
    "            Nt = np.linalg.det(N)\n",
    "            a[j] = (step*Nt)\n",
    "\n",
    "        x1 = cp.Variable(n)\n",
    "        obj1 = cp.Maximize(a * x1)\n",
    "        one = np.ones((p, 1))\n",
    "        constraints1 = [B * x1 >= 0, (np.dot(one.T, B)) * x1 == 1]\n",
    "        prob1 = cp.Problem(obj1, constraints1)\n",
    "        prob1.solve(solver=cp.CVXOPT)\n",
    "        a1 = prob1.value\n",
    "        x1 = x1.value\n",
    "\n",
    "        x2 = cp.Variable(n)\n",
    "        obj2 = cp.Minimize(a * x2)\n",
    "        constraints2 = [B * x2 >= 0, (np.dot(one.T, B)) * x2 == 1]\n",
    "        prob2 = cp.Problem(obj2, constraints2)\n",
    "        prob2.solve(solver=cp.CVXOPT)\n",
    "        a2 = prob2.value\n",
    "        x2 = x2.value\n",
    "\n",
    "        if abs(a1) > abs(a2):\n",
    "            M[:, i] = x1;\n",
    "        else:\n",
    "            M[:, i] = x2;\n",
    "\n",
    "        detM = np.linalg.det(M)\n",
    "        conv.append(abs(detM))\n",
    "        mm += 1\n",
    "\n",
    "    # Convergence condition\n",
    "    idx = len(conv) - 1\n",
    "    if (mm > 1) and (abs(conv[idx] - conv[idx-1]) < 1e-6):\n",
    "        break  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "C = np.dot(B, M)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(425, 10)"
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "C.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "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>Topic 1 words</th>\n",
       "      <th>Topic 1 weights</th>\n",
       "      <th>Topic 2 words</th>\n",
       "      <th>Topic 2 weights</th>\n",
       "      <th>Topic 3 words</th>\n",
       "      <th>Topic 3 weights</th>\n",
       "      <th>Topic 4 words</th>\n",
       "      <th>Topic 4 weights</th>\n",
       "      <th>Topic 5 words</th>\n",
       "      <th>Topic 5 weights</th>\n",
       "      <th>Topic 6 words</th>\n",
       "      <th>Topic 6 weights</th>\n",
       "      <th>Topic 7 words</th>\n",
       "      <th>Topic 7 weights</th>\n",
       "      <th>Topic 8 words</th>\n",
       "      <th>Topic 8 weights</th>\n",
       "      <th>Topic 9 words</th>\n",
       "      <th>Topic 9 weights</th>\n",
       "      <th>Topic 10 words</th>\n",
       "      <th>Topic 10 weights</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>collagen</td>\n",
       "      <td>0.1034</td>\n",
       "      <td>rfk</td>\n",
       "      <td>0.0544</td>\n",
       "      <td>limbo</td>\n",
       "      <td>0.0480</td>\n",
       "      <td>matter</td>\n",
       "      <td>0.0462</td>\n",
       "      <td>habitat</td>\n",
       "      <td>0.0552</td>\n",
       "      <td>safeti</td>\n",
       "      <td>0.0679</td>\n",
       "      <td>warm</td>\n",
       "      <td>0.0186</td>\n",
       "      <td>climat</td>\n",
       "      <td>0.0204</td>\n",
       "      <td>climat</td>\n",
       "      <td>0.0207</td>\n",
       "      <td>volcan</td>\n",
       "      <td>0.0179</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>cliniqu</td>\n",
       "      <td>0.1031</td>\n",
       "      <td>jr</td>\n",
       "      <td>0.0520</td>\n",
       "      <td>exit</td>\n",
       "      <td>0.0478</td>\n",
       "      <td>money</td>\n",
       "      <td>0.0430</td>\n",
       "      <td>wildlif</td>\n",
       "      <td>0.0511</td>\n",
       "      <td>silli</td>\n",
       "      <td>0.0666</td>\n",
       "      <td>global</td>\n",
       "      <td>0.0186</td>\n",
       "      <td>chang</td>\n",
       "      <td>0.0202</td>\n",
       "      <td>chang</td>\n",
       "      <td>0.0207</td>\n",
       "      <td>erupt</td>\n",
       "      <td>0.0168</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>scream</td>\n",
       "      <td>0.1030</td>\n",
       "      <td>month</td>\n",
       "      <td>0.0459</td>\n",
       "      <td>bill</td>\n",
       "      <td>0.0429</td>\n",
       "      <td>resourc</td>\n",
       "      <td>0.0413</td>\n",
       "      <td>protect</td>\n",
       "      <td>0.0460</td>\n",
       "      <td>mine</td>\n",
       "      <td>0.0592</td>\n",
       "      <td>exit</td>\n",
       "      <td>0.0177</td>\n",
       "      <td>agenc</td>\n",
       "      <td>0.0197</td>\n",
       "      <td>safeti</td>\n",
       "      <td>0.0154</td>\n",
       "      <td>senat</td>\n",
       "      <td>0.0158</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>clinic</td>\n",
       "      <td>0.1003</td>\n",
       "      <td>ago</td>\n",
       "      <td>0.0445</td>\n",
       "      <td>graham</td>\n",
       "      <td>0.0378</td>\n",
       "      <td>better</td>\n",
       "      <td>0.0399</td>\n",
       "      <td>action</td>\n",
       "      <td>0.0354</td>\n",
       "      <td>regul</td>\n",
       "      <td>0.0393</td>\n",
       "      <td>limbo</td>\n",
       "      <td>0.0176</td>\n",
       "      <td>new</td>\n",
       "      <td>0.0184</td>\n",
       "      <td>silli</td>\n",
       "      <td>0.0151</td>\n",
       "      <td>volcano</td>\n",
       "      <td>0.0155</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>trial</td>\n",
       "      <td>0.1003</td>\n",
       "      <td>mean</td>\n",
       "      <td>0.0316</td>\n",
       "      <td>put</td>\n",
       "      <td>0.0360</td>\n",
       "      <td>part</td>\n",
       "      <td>0.0343</td>\n",
       "      <td>help</td>\n",
       "      <td>0.0350</td>\n",
       "      <td>call</td>\n",
       "      <td>0.0301</td>\n",
       "      <td>put</td>\n",
       "      <td>0.0143</td>\n",
       "      <td>form</td>\n",
       "      <td>0.0162</td>\n",
       "      <td>mine</td>\n",
       "      <td>0.0142</td>\n",
       "      <td>iceland</td>\n",
       "      <td>0.0150</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>clip</td>\n",
       "      <td>0.0969</td>\n",
       "      <td>cold</td>\n",
       "      <td>0.0297</td>\n",
       "      <td>climat</td>\n",
       "      <td>0.0221</td>\n",
       "      <td>save</td>\n",
       "      <td>0.0341</td>\n",
       "      <td>#saveterra</td>\n",
       "      <td>0.0319</td>\n",
       "      <td>coal</td>\n",
       "      <td>0.0257</td>\n",
       "      <td>graham</td>\n",
       "      <td>0.0134</td>\n",
       "      <td>feder</td>\n",
       "      <td>0.0158</td>\n",
       "      <td>senat</td>\n",
       "      <td>0.0101</td>\n",
       "      <td>climat</td>\n",
       "      <td>0.0148</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>art</td>\n",
       "      <td>0.0562</td>\n",
       "      <td>dc</td>\n",
       "      <td>0.0286</td>\n",
       "      <td>unveil</td>\n",
       "      <td>0.0217</td>\n",
       "      <td>live</td>\n",
       "      <td>0.0312</td>\n",
       "      <td>take</td>\n",
       "      <td>0.0302</td>\n",
       "      <td>global</td>\n",
       "      <td>0.0181</td>\n",
       "      <td>bill</td>\n",
       "      <td>0.0126</td>\n",
       "      <td>administr</td>\n",
       "      <td>0.0122</td>\n",
       "      <td>regul</td>\n",
       "      <td>0.0101</td>\n",
       "      <td>chang</td>\n",
       "      <td>0.0148</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>chang</td>\n",
       "      <td>0.0150</td>\n",
       "      <td>snow</td>\n",
       "      <td>0.0247</td>\n",
       "      <td>chang</td>\n",
       "      <td>0.0217</td>\n",
       "      <td>believ</td>\n",
       "      <td>0.0256</td>\n",
       "      <td>#climatechange</td>\n",
       "      <td>0.0259</td>\n",
       "      <td>warm</td>\n",
       "      <td>0.0179</td>\n",
       "      <td>climat</td>\n",
       "      <td>0.0108</td>\n",
       "      <td>propos</td>\n",
       "      <td>0.0112</td>\n",
       "      <td>via</td>\n",
       "      <td>0.0098</td>\n",
       "      <td>trigger</td>\n",
       "      <td>0.0143</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>climat</td>\n",
       "      <td>0.0148</td>\n",
       "      <td>global</td>\n",
       "      <td>0.0153</td>\n",
       "      <td>senat</td>\n",
       "      <td>0.0216</td>\n",
       "      <td>global</td>\n",
       "      <td>0.0153</td>\n",
       "      <td>chang</td>\n",
       "      <td>0.0175</td>\n",
       "      <td>effort</td>\n",
       "      <td>0.0150</td>\n",
       "      <td>chang</td>\n",
       "      <td>0.0107</td>\n",
       "      <td>obama</td>\n",
       "      <td>0.0102</td>\n",
       "      <td>peopl</td>\n",
       "      <td>0.0095</td>\n",
       "      <td>compromis</td>\n",
       "      <td>0.0138</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>keep</td>\n",
       "      <td>0.0123</td>\n",
       "      <td>warm</td>\n",
       "      <td>0.0153</td>\n",
       "      <td>talk</td>\n",
       "      <td>0.0183</td>\n",
       "      <td>warm</td>\n",
       "      <td>0.0153</td>\n",
       "      <td>climat</td>\n",
       "      <td>0.0172</td>\n",
       "      <td>well</td>\n",
       "      <td>0.0126</td>\n",
       "      <td>gore</td>\n",
       "      <td>0.0086</td>\n",
       "      <td>studi</td>\n",
       "      <td>0.0099</td>\n",
       "      <td>world</td>\n",
       "      <td>0.0090</td>\n",
       "      <td>global</td>\n",
       "      <td>0.0133</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  Topic 1 words Topic 1 weights Topic 2 words Topic 2 weights Topic 3 words  \\\n",
       "0      collagen          0.1034           rfk          0.0544         limbo   \n",
       "1       cliniqu          0.1031            jr          0.0520          exit   \n",
       "2        scream          0.1030         month          0.0459          bill   \n",
       "3        clinic          0.1003           ago          0.0445        graham   \n",
       "4         trial          0.1003          mean          0.0316           put   \n",
       "5          clip          0.0969          cold          0.0297        climat   \n",
       "6           art          0.0562            dc          0.0286        unveil   \n",
       "7         chang          0.0150          snow          0.0247         chang   \n",
       "8        climat          0.0148        global          0.0153         senat   \n",
       "9          keep          0.0123          warm          0.0153          talk   \n",
       "\n",
       "  Topic 3 weights Topic 4 words Topic 4 weights   Topic 5 words  \\\n",
       "0          0.0480        matter          0.0462         habitat   \n",
       "1          0.0478         money          0.0430         wildlif   \n",
       "2          0.0429       resourc          0.0413         protect   \n",
       "3          0.0378        better          0.0399          action   \n",
       "4          0.0360          part          0.0343            help   \n",
       "5          0.0221          save          0.0341      #saveterra   \n",
       "6          0.0217          live          0.0312            take   \n",
       "7          0.0217        believ          0.0256  #climatechange   \n",
       "8          0.0216        global          0.0153           chang   \n",
       "9          0.0183          warm          0.0153          climat   \n",
       "\n",
       "  Topic 5 weights Topic 6 words Topic 6 weights Topic 7 words Topic 7 weights  \\\n",
       "0          0.0552        safeti          0.0679          warm          0.0186   \n",
       "1          0.0511         silli          0.0666        global          0.0186   \n",
       "2          0.0460          mine          0.0592          exit          0.0177   \n",
       "3          0.0354         regul          0.0393         limbo          0.0176   \n",
       "4          0.0350          call          0.0301           put          0.0143   \n",
       "5          0.0319          coal          0.0257        graham          0.0134   \n",
       "6          0.0302        global          0.0181          bill          0.0126   \n",
       "7          0.0259          warm          0.0179        climat          0.0108   \n",
       "8          0.0175        effort          0.0150         chang          0.0107   \n",
       "9          0.0172          well          0.0126          gore          0.0086   \n",
       "\n",
       "  Topic 8 words Topic 8 weights Topic 9 words Topic 9 weights Topic 10 words  \\\n",
       "0        climat          0.0204        climat          0.0207         volcan   \n",
       "1         chang          0.0202         chang          0.0207          erupt   \n",
       "2         agenc          0.0197        safeti          0.0154          senat   \n",
       "3           new          0.0184         silli          0.0151        volcano   \n",
       "4          form          0.0162          mine          0.0142        iceland   \n",
       "5         feder          0.0158         senat          0.0101         climat   \n",
       "6     administr          0.0122         regul          0.0101          chang   \n",
       "7        propos          0.0112           via          0.0098        trigger   \n",
       "8         obama          0.0102         peopl          0.0095      compromis   \n",
       "9         studi          0.0099         world          0.0090         global   \n",
       "\n",
       "  Topic 10 weights  \n",
       "0           0.0179  \n",
       "1           0.0168  \n",
       "2           0.0158  \n",
       "3           0.0155  \n",
       "4           0.0150  \n",
       "5           0.0148  \n",
       "6           0.0148  \n",
       "7           0.0143  \n",
       "8           0.0138  \n",
       "9           0.0133  "
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "no_top_words = 10\n",
    "topic_dict_AF = {}\n",
    "for k in range(F):\n",
    "    topic_dict_AF[\"Topic %d words\" % (k+1)]= ['{}'.format(tfidf_feature_names[i]) for i in C[:, k].argsort()[:-no_top_words - 1:-1]]\n",
    "    topic_dict_AF[\"Topic %d weights\" % (k+1)]= ['{:.4f}'.format(C[:, k][i]) for i in C[:, k].argsort()[:-no_top_words - 1:-1]]    \n",
    "pd.DataFrame(topic_dict_AF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "table_AF = pd.DataFrame(topic_dict_AF)\n",
    "table_AF.to_csv(\"AF.csv\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## kmeans clustering ##"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 353 ms\n"
     ]
    }
   ],
   "source": [
    "from sklearn.cluster import KMeans\n",
    "\n",
    "F = 10 # num of topics\n",
    "model_kmeans = KMeans(n_clusters=F, init='k-means++', max_iter=100, n_init=1)\n",
    "%time model_kmeans.fit(tfidf)\n",
    "\n",
    "order_centroids = model_kmeans.cluster_centers_.argsort()[:, ::-1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "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>Topic 1 words</th>\n",
       "      <th>Topic 1 weights</th>\n",
       "      <th>Topic 2 words</th>\n",
       "      <th>Topic 2 weights</th>\n",
       "      <th>Topic 3 words</th>\n",
       "      <th>Topic 3 weights</th>\n",
       "      <th>Topic 4 words</th>\n",
       "      <th>Topic 4 weights</th>\n",
       "      <th>Topic 5 words</th>\n",
       "      <th>Topic 5 weights</th>\n",
       "      <th>Topic 6 words</th>\n",
       "      <th>Topic 6 weights</th>\n",
       "      <th>Topic 7 words</th>\n",
       "      <th>Topic 7 weights</th>\n",
       "      <th>Topic 8 words</th>\n",
       "      <th>Topic 8 weights</th>\n",
       "      <th>Topic 9 words</th>\n",
       "      <th>Topic 9 weights</th>\n",
       "      <th>Topic 10 words</th>\n",
       "      <th>Topic 10 weights</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>scientist</td>\n",
       "      <td>0.0423</td>\n",
       "      <td>rfk</td>\n",
       "      <td>0.1701</td>\n",
       "      <td>#tcot</td>\n",
       "      <td>0.0549</td>\n",
       "      <td>trial</td>\n",
       "      <td>0.1189</td>\n",
       "      <td>wors</td>\n",
       "      <td>0.1007</td>\n",
       "      <td>york</td>\n",
       "      <td>0.1662</td>\n",
       "      <td>chang</td>\n",
       "      <td>0.0147</td>\n",
       "      <td>warm</td>\n",
       "      <td>0.0145</td>\n",
       "      <td>ocean</td>\n",
       "      <td>0.1206</td>\n",
       "      <td>#globalwarming</td>\n",
       "      <td>0.2008</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>scienc</td>\n",
       "      <td>0.0356</td>\n",
       "      <td>jr</td>\n",
       "      <td>0.1637</td>\n",
       "      <td>#teaparty</td>\n",
       "      <td>0.0278</td>\n",
       "      <td>clinic</td>\n",
       "      <td>0.1189</td>\n",
       "      <td>allergi</td>\n",
       "      <td>0.0552</td>\n",
       "      <td>new</td>\n",
       "      <td>0.0532</td>\n",
       "      <td>climat</td>\n",
       "      <td>0.0145</td>\n",
       "      <td>global</td>\n",
       "      <td>0.0145</td>\n",
       "      <td>intensifi</td>\n",
       "      <td>0.0322</td>\n",
       "      <td>#saveterra</td>\n",
       "      <td>0.1051</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>debat</td>\n",
       "      <td>0.0173</td>\n",
       "      <td>month</td>\n",
       "      <td>0.1161</td>\n",
       "      <td>#gop</td>\n",
       "      <td>0.0248</td>\n",
       "      <td>collagen</td>\n",
       "      <td>0.1133</td>\n",
       "      <td>season</td>\n",
       "      <td>0.0401</td>\n",
       "      <td>fall</td>\n",
       "      <td>0.0217</td>\n",
       "      <td>via</td>\n",
       "      <td>0.0066</td>\n",
       "      <td>snow</td>\n",
       "      <td>0.0046</td>\n",
       "      <td>show</td>\n",
       "      <td>0.0232</td>\n",
       "      <td>fact</td>\n",
       "      <td>0.0772</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>art</td>\n",
       "      <td>0.0095</td>\n",
       "      <td>ago</td>\n",
       "      <td>0.1131</td>\n",
       "      <td>#tlot</td>\n",
       "      <td>0.0228</td>\n",
       "      <td>scream</td>\n",
       "      <td>0.0927</td>\n",
       "      <td>make</td>\n",
       "      <td>0.0370</td>\n",
       "      <td>citi</td>\n",
       "      <td>0.0215</td>\n",
       "      <td>bill</td>\n",
       "      <td>0.0048</td>\n",
       "      <td>gore</td>\n",
       "      <td>0.0035</td>\n",
       "      <td>rise</td>\n",
       "      <td>0.0228</td>\n",
       "      <td>tell</td>\n",
       "      <td>0.0765</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>climat</td>\n",
       "      <td>0.0093</td>\n",
       "      <td>mean</td>\n",
       "      <td>0.0667</td>\n",
       "      <td>#ocra</td>\n",
       "      <td>0.0215</td>\n",
       "      <td>cliniqu</td>\n",
       "      <td>0.0869</td>\n",
       "      <td>blizzard</td>\n",
       "      <td>0.0360</td>\n",
       "      <td>commun</td>\n",
       "      <td>0.0154</td>\n",
       "      <td>senat</td>\n",
       "      <td>0.0036</td>\n",
       "      <td>volcano</td>\n",
       "      <td>0.0035</td>\n",
       "      <td>faster</td>\n",
       "      <td>0.0208</td>\n",
       "      <td>get</td>\n",
       "      <td>0.0407</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>warm</td>\n",
       "      <td>0.0091</td>\n",
       "      <td>cold</td>\n",
       "      <td>0.0636</td>\n",
       "      <td>disprov</td>\n",
       "      <td>0.0207</td>\n",
       "      <td>clip</td>\n",
       "      <td>0.0781</td>\n",
       "      <td>snowstorm</td>\n",
       "      <td>0.0253</td>\n",
       "      <td>weather</td>\n",
       "      <td>0.0147</td>\n",
       "      <td>energi</td>\n",
       "      <td>0.0036</td>\n",
       "      <td>caus</td>\n",
       "      <td>0.0034</td>\n",
       "      <td>water</td>\n",
       "      <td>0.0151</td>\n",
       "      <td>movement</td>\n",
       "      <td>0.0155</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>global</td>\n",
       "      <td>0.0090</td>\n",
       "      <td>dc</td>\n",
       "      <td>0.0558</td>\n",
       "      <td>#p</td>\n",
       "      <td>0.0181</td>\n",
       "      <td>art</td>\n",
       "      <td>0.0396</td>\n",
       "      <td>ever</td>\n",
       "      <td>0.0232</td>\n",
       "      <td>check</td>\n",
       "      <td>0.0143</td>\n",
       "      <td>u</td>\n",
       "      <td>0.0034</td>\n",
       "      <td>believ</td>\n",
       "      <td>0.0033</td>\n",
       "      <td>impact</td>\n",
       "      <td>0.0147</td>\n",
       "      <td>support</td>\n",
       "      <td>0.0137</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>chang</td>\n",
       "      <td>0.0068</td>\n",
       "      <td>snow</td>\n",
       "      <td>0.0362</td>\n",
       "      <td>#climategate</td>\n",
       "      <td>0.0117</td>\n",
       "      <td>chang</td>\n",
       "      <td>0.0104</td>\n",
       "      <td>time</td>\n",
       "      <td>0.0226</td>\n",
       "      <td>big</td>\n",
       "      <td>0.0116</td>\n",
       "      <td>new</td>\n",
       "      <td>0.0034</td>\n",
       "      <td>al</td>\n",
       "      <td>0.0031</td>\n",
       "      <td>carbon</td>\n",
       "      <td>0.0125</td>\n",
       "      <td>next</td>\n",
       "      <td>0.0107</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>say</td>\n",
       "      <td>0.0059</td>\n",
       "      <td>f</td>\n",
       "      <td>0.0161</td>\n",
       "      <td>warm</td>\n",
       "      <td>0.0094</td>\n",
       "      <td>climat</td>\n",
       "      <td>0.0102</td>\n",
       "      <td>c</td>\n",
       "      <td>0.0223</td>\n",
       "      <td>scam</td>\n",
       "      <td>0.0114</td>\n",
       "      <td>fight</td>\n",
       "      <td>0.0034</td>\n",
       "      <td>could</td>\n",
       "      <td>0.0030</td>\n",
       "      <td>nation</td>\n",
       "      <td>0.0117</td>\n",
       "      <td>say</td>\n",
       "      <td>0.0076</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>heat</td>\n",
       "      <td>0.0056</td>\n",
       "      <td>washington</td>\n",
       "      <td>0.0141</td>\n",
       "      <td>global</td>\n",
       "      <td>0.0093</td>\n",
       "      <td>day</td>\n",
       "      <td>0.0099</td>\n",
       "      <td>blame</td>\n",
       "      <td>0.0185</td>\n",
       "      <td>go</td>\n",
       "      <td>0.0096</td>\n",
       "      <td>immigr</td>\n",
       "      <td>0.0034</td>\n",
       "      <td>storm</td>\n",
       "      <td>0.0030</td>\n",
       "      <td>year</td>\n",
       "      <td>0.0115</td>\n",
       "      <td>chang</td>\n",
       "      <td>0.0068</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  Topic 1 words Topic 1 weights Topic 2 words Topic 2 weights Topic 3 words  \\\n",
       "0     scientist          0.0423           rfk          0.1701         #tcot   \n",
       "1        scienc          0.0356            jr          0.1637     #teaparty   \n",
       "2         debat          0.0173         month          0.1161          #gop   \n",
       "3           art          0.0095           ago          0.1131         #tlot   \n",
       "4        climat          0.0093          mean          0.0667         #ocra   \n",
       "5          warm          0.0091          cold          0.0636       disprov   \n",
       "6        global          0.0090            dc          0.0558            #p   \n",
       "7         chang          0.0068          snow          0.0362  #climategate   \n",
       "8           say          0.0059             f          0.0161          warm   \n",
       "9          heat          0.0056    washington          0.0141        global   \n",
       "\n",
       "  Topic 3 weights Topic 4 words Topic 4 weights Topic 5 words Topic 5 weights  \\\n",
       "0          0.0549         trial          0.1189          wors          0.1007   \n",
       "1          0.0278        clinic          0.1189       allergi          0.0552   \n",
       "2          0.0248      collagen          0.1133        season          0.0401   \n",
       "3          0.0228        scream          0.0927          make          0.0370   \n",
       "4          0.0215       cliniqu          0.0869      blizzard          0.0360   \n",
       "5          0.0207          clip          0.0781     snowstorm          0.0253   \n",
       "6          0.0181           art          0.0396          ever          0.0232   \n",
       "7          0.0117         chang          0.0104          time          0.0226   \n",
       "8          0.0094        climat          0.0102             c          0.0223   \n",
       "9          0.0093           day          0.0099         blame          0.0185   \n",
       "\n",
       "  Topic 6 words Topic 6 weights Topic 7 words Topic 7 weights Topic 8 words  \\\n",
       "0          york          0.1662         chang          0.0147          warm   \n",
       "1           new          0.0532        climat          0.0145        global   \n",
       "2          fall          0.0217           via          0.0066          snow   \n",
       "3          citi          0.0215          bill          0.0048          gore   \n",
       "4        commun          0.0154         senat          0.0036       volcano   \n",
       "5       weather          0.0147        energi          0.0036          caus   \n",
       "6         check          0.0143             u          0.0034        believ   \n",
       "7           big          0.0116           new          0.0034            al   \n",
       "8          scam          0.0114         fight          0.0034         could   \n",
       "9            go          0.0096        immigr          0.0034         storm   \n",
       "\n",
       "  Topic 8 weights Topic 9 words Topic 9 weights  Topic 10 words  \\\n",
       "0          0.0145         ocean          0.1206  #globalwarming   \n",
       "1          0.0145     intensifi          0.0322      #saveterra   \n",
       "2          0.0046          show          0.0232            fact   \n",
       "3          0.0035          rise          0.0228            tell   \n",
       "4          0.0035        faster          0.0208             get   \n",
       "5          0.0034         water          0.0151        movement   \n",
       "6          0.0033        impact          0.0147         support   \n",
       "7          0.0031        carbon          0.0125            next   \n",
       "8          0.0030        nation          0.0117             say   \n",
       "9          0.0030          year          0.0115           chang   \n",
       "\n",
       "  Topic 10 weights  \n",
       "0           0.2008  \n",
       "1           0.1051  \n",
       "2           0.0772  \n",
       "3           0.0765  \n",
       "4           0.0407  \n",
       "5           0.0155  \n",
       "6           0.0137  \n",
       "7           0.0107  \n",
       "8           0.0076  \n",
       "9           0.0068  "
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "no_top_words = 10\n",
    "topic_dict_kmeans = {}\n",
    "for k in range(F):\n",
    "    topic_dict_kmeans[\"Topic %d words\" % (k+1)] = ['{}'.format(tfidf_feature_names[i]) for i in order_centroids[k, :no_top_words]]\n",
    "    topic_dict_kmeans[\"Topic %d weights\" % (k+1)]= ['{:.4f}'.format(model_kmeans.cluster_centers_.T[:, k][i]) for i in order_centroids[k, :no_top_words]]    \n",
    "pd.DataFrame(topic_dict_kmeans)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [],
   "source": [
    "table_kmeans = pd.DataFrame(topic_dict_kmeans)\n",
    "table_kmeans.to_csv(\"kmeans.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}