EDA.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* read the texts\n",
    "* format them\n",
    "* lemmatize\n",
    "* stem \n",
    "* clear the garbage\n",
    "* do tf-idf\n",
    "* do K-Means"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.cm as cm\n",
    "\n",
    "from sklearn.cluster import MiniBatchKMeans\n",
    "from sklearn.cluster import KMeans\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "from sklearn.decomposition import PCA\n",
    "from sklearn.manifold import TSNE\n",
    "from nltk.stem.porter import PorterStemmer\n",
    "\n",
    "import nltk"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[nltk_data] Downloading package punkt to /home/lodya/nltk_data...\n",
      "[nltk_data]   Unzipping tokenizers/punkt.zip.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nltk.download('punkt')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "from pathlib import Path"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.DataFrame(columns=('id', 'text'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "import re\n",
    "from tqdm import tqdm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 2769/2769 [00:00<00:00, 4284.81it/s]\n"
     ]
    }
   ],
   "source": [
    "dirpath = '/home/lodya/Desktop/Projects/Term_Project_1/subs/plain'\n",
    "data = []\n",
    "for filename in tqdm(os.listdir(dirpath)):\n",
    "    with open(Path(dirpath, filename)) as f:\n",
    "        text_id = filename.split('.')[0]\n",
    "        text = f.read()\n",
    "        text = re.sub('\\[.*\\]|-', '', text)\n",
    "        data.append((text_id, text))\n",
    "df = pd.DataFrame(data=data, columns=('id', 'text'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "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>id</th>\n",
       "      <th>text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>cSw5R-jdMiI</td>\n",
       "      <td>Let g of x equal one over x. Can we use the ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ieC5LgoGJVg</td>\n",
       "      <td>So we've got the expression two X plus four ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>AhnJtTI_DMM</td>\n",
       "      <td>Let's find the limit of f of x times h of x ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>O_yeKo6-qIg</td>\n",
       "      <td>Anytime you're trying to come up with a mech...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>HiY5UElqfhs</td>\n",
       "      <td>If a sub n is equal to n squared minus 10 ov...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            id                                               text\n",
       "0  cSw5R-jdMiI    Let g of x equal one over x. Can we use the ...\n",
       "1  ieC5LgoGJVg    So we've got the expression two X plus four ...\n",
       "2  AhnJtTI_DMM    Let's find the limit of f of x times h of x ...\n",
       "3  O_yeKo6-qIg    Anytime you're trying to come up with a mech...\n",
       "4  HiY5UElqfhs    If a sub n is equal to n squared minus 10 ov..."
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "# remove bracketed comments at the beggining of the texts"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "stemmer = PorterStemmer()\n",
    "\n",
    "def stem_words(words_list, stemmer):\n",
    "    return [stemmer.stem(word) for word in words_list]\n",
    "\n",
    "def tokenize(text):\n",
    "    tokens = nltk.word_tokenize(text)\n",
    "    stems = stem_words(tokens, stemmer)\n",
    "    return stems"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [],
   "source": [
    "tf_idf = TfidfVectorizer(stop_words='english')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [],
   "source": [
    "matrix = tf_idf.fit_transform(df['text'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "  0%|          | 0/99 [00:00<?, ?it/s]\u001b[A\n",
      "  1%|          | 1/99 [00:00<00:10,  9.72it/s]\u001b[A\n",
      "  2%|▏         | 2/99 [00:00<00:10,  9.43it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 2 clusters\n",
      "Fit 3 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "  3%|▎         | 3/99 [00:00<00:10,  9.51it/s]\u001b[A\n",
      "  4%|▍         | 4/99 [00:00<00:11,  8.28it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 4 clusters\n",
      "Fit 5 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "  5%|▌         | 5/99 [00:00<00:14,  6.34it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 6 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "  6%|▌         | 6/99 [00:00<00:17,  5.29it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 7 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "  7%|▋         | 7/99 [00:01<00:21,  4.26it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 8 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "  8%|▊         | 8/99 [00:01<00:29,  3.04it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 9 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "  9%|▉         | 9/99 [00:02<00:34,  2.65it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 10 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 10%|█         | 10/99 [00:02<00:36,  2.42it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 11 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 11%|█         | 11/99 [00:03<00:38,  2.28it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 12 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 12%|█▏        | 12/99 [00:04<00:44,  1.96it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 13 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 13%|█▎        | 13/99 [00:04<00:55,  1.56it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 14 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 14%|█▍        | 14/99 [00:05<00:56,  1.51it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 15 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 15%|█▌        | 15/99 [00:06<00:57,  1.47it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 16 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 16%|█▌        | 16/99 [00:07<00:58,  1.42it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 17 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 17%|█▋        | 17/99 [00:07<00:59,  1.37it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 18 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 18%|█▊        | 18/99 [00:08<01:01,  1.31it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 19 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 19%|█▉        | 19/99 [00:09<01:04,  1.23it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 20 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 20%|██        | 20/99 [00:11<01:16,  1.03it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 21 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 21%|██        | 21/99 [00:12<01:27,  1.12s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 22 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 22%|██▏       | 22/99 [00:14<01:34,  1.23s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 23 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 23%|██▎       | 23/99 [00:15<01:35,  1.25s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 24 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 24%|██▍       | 24/99 [00:16<01:39,  1.33s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 25 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 25%|██▌       | 25/99 [00:18<01:35,  1.30s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 26 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 26%|██▋       | 26/99 [00:19<01:37,  1.34s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 27 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 27%|██▋       | 27/99 [00:21<01:47,  1.49s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 28 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 28%|██▊       | 28/99 [00:22<01:46,  1.50s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 29 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 29%|██▉       | 29/99 [00:24<01:57,  1.68s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 30 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 30%|███       | 30/99 [00:26<01:55,  1.67s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 31 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 31%|███▏      | 31/99 [00:28<01:53,  1.67s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 32 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 32%|███▏      | 32/99 [00:29<01:49,  1.63s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 33 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 33%|███▎      | 33/99 [00:31<01:44,  1.58s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 34 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 34%|███▍      | 34/99 [00:33<01:52,  1.73s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 35 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 35%|███▌      | 35/99 [00:35<01:59,  1.86s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 36 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 36%|███▋      | 36/99 [00:38<02:10,  2.07s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 37 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 37%|███▋      | 37/99 [00:40<02:15,  2.18s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 38 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 38%|███▊      | 38/99 [00:43<02:22,  2.34s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 39 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 39%|███▉      | 39/99 [00:45<02:10,  2.18s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 40 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 40%|████      | 40/99 [00:46<02:02,  2.08s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 41 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 41%|████▏     | 41/99 [00:48<01:56,  2.01s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 42 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 42%|████▏     | 42/99 [00:50<01:54,  2.01s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 43 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 43%|████▎     | 43/99 [00:53<02:09,  2.31s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 44 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 44%|████▍     | 44/99 [00:57<02:27,  2.68s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 45 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 45%|████▌     | 45/99 [01:00<02:27,  2.73s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 46 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 46%|████▋     | 46/99 [01:02<02:15,  2.55s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 47 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 47%|████▋     | 47/99 [01:04<02:06,  2.42s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 48 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 48%|████▊     | 48/99 [01:07<02:11,  2.58s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 49 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 49%|████▉     | 49/99 [01:10<02:14,  2.69s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 50 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 51%|█████     | 50/99 [01:12<02:11,  2.69s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 51 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 52%|█████▏    | 51/99 [01:15<02:08,  2.69s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 52 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 53%|█████▎    | 52/99 [01:18<02:04,  2.65s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 53 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 54%|█████▎    | 53/99 [01:20<02:00,  2.63s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 54 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 55%|█████▍    | 54/99 [01:24<02:14,  2.99s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 55 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 56%|█████▌    | 55/99 [01:28<02:22,  3.23s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 56 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 57%|█████▋    | 56/99 [01:31<02:18,  3.22s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 57 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 58%|█████▊    | 57/99 [01:35<02:22,  3.39s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 58 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 59%|█████▊    | 58/99 [01:38<02:10,  3.17s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 59 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 60%|█████▉    | 59/99 [01:41<02:06,  3.17s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 60 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 61%|██████    | 60/99 [01:44<02:01,  3.11s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 61 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 62%|██████▏   | 61/99 [01:47<01:54,  3.02s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 62 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 63%|██████▎   | 62/99 [01:50<02:01,  3.29s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 63 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 64%|██████▎   | 63/99 [01:54<01:58,  3.30s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 64 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 65%|██████▍   | 64/99 [01:58<02:01,  3.49s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 65 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 66%|██████▌   | 65/99 [02:02<02:04,  3.67s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 66 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 67%|██████▋   | 66/99 [02:05<01:57,  3.56s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 67 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 68%|██████▊   | 67/99 [02:08<01:49,  3.42s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 68 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 69%|██████▊   | 68/99 [02:11<01:43,  3.34s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 69 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 70%|██████▉   | 69/99 [02:15<01:42,  3.42s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 70 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 71%|███████   | 70/99 [02:18<01:38,  3.38s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 71 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 72%|███████▏  | 71/99 [02:23<01:43,  3.69s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 72 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 73%|███████▎  | 72/99 [02:26<01:38,  3.64s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 73 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 74%|███████▎  | 73/99 [02:30<01:33,  3.59s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 74 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 75%|███████▍  | 74/99 [02:33<01:30,  3.61s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 75 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 76%|███████▌  | 75/99 [02:38<01:36,  4.03s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 76 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 77%|███████▋  | 76/99 [02:44<01:41,  4.41s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 77 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 78%|███████▊  | 77/99 [02:49<01:46,  4.84s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 78 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 79%|███████▉  | 78/99 [02:56<01:51,  5.29s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 79 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 80%|███████▉  | 79/99 [03:00<01:39,  4.97s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 80 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 81%|████████  | 80/99 [03:05<01:35,  5.02s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 81 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 82%|████████▏ | 81/99 [03:10<01:27,  4.85s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 82 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 83%|████████▎ | 82/99 [03:15<01:24,  4.99s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 83 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 84%|████████▍ | 83/99 [03:19<01:14,  4.68s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 84 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 85%|████████▍ | 84/99 [03:25<01:16,  5.09s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 85 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 86%|████████▌ | 85/99 [03:29<01:06,  4.76s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 86 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 87%|████████▋ | 86/99 [03:33<00:58,  4.53s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 87 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 88%|████████▊ | 87/99 [03:37<00:53,  4.44s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 88 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 89%|████████▉ | 88/99 [03:44<00:57,  5.26s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 89 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 90%|████████▉ | 89/99 [03:49<00:51,  5.10s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 90 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 91%|█████████ | 90/99 [03:53<00:43,  4.84s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 91 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 92%|█████████▏| 91/99 [03:59<00:40,  5.07s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 92 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 93%|█████████▎| 92/99 [04:05<00:38,  5.48s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 93 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 94%|█████████▍| 93/99 [04:11<00:32,  5.50s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 94 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 95%|█████████▍| 94/99 [04:15<00:25,  5.18s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 95 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 96%|█████████▌| 95/99 [04:20<00:20,  5.07s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 96 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 97%|█████████▋| 96/99 [04:26<00:16,  5.34s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 97 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 98%|█████████▊| 97/99 [04:31<00:10,  5.25s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 98 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      " 99%|█████████▉| 98/99 [04:37<00:05,  5.53s/it]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 99 clusters\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "100%|██████████| 99/99 [04:43<00:00,  5.69s/it]\u001b[A\n",
      "\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fit 100 clusters\n"
     ]
    },
    {
     "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": [
    "def find_optimal_clusters(data, max_k):\n",
    "    iters = range(2, max_k+1, 1)\n",
    "    \n",
    "    sse = []\n",
    "    for k in tqdm(iters):\n",
    "        sse.append(MiniBatchKMeans(n_clusters=k, batch_size=512,random_state=27, verbose=0).fit(data).inertia_)\n",
    "        print('Fit {} clusters'.format(k))\n",
    "        \n",
    "    f, ax = plt.subplots(1, 1)\n",
    "    ax.plot(iters, sse, marker='o')\n",
    "    ax.set_xlabel('Cluster Centers')\n",
    "    ax.set_xticks(iters)\n",
    "    ax.set_xticklabels(iters)\n",
    "    ax.set_ylabel('SSE')\n",
    "    ax.set_title('SSE by Cluster Center Plot')\n",
    "    \n",
    "find_optimal_clusters(matrix, 100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [],
   "source": [
    "clusters = MiniBatchKMeans(n_clusters=15, random_state=27, verbose=0).fit_predict(matrix)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2769,)"
      ]
     },
     "execution_count": 75,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clusters.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Cluster 0\n",
      "let,ohms,r1,response,resistance,inductor,plus,zero,r2,minus,node,capacitor,gonna,resistors,volts,equals,resistor,circuit,current,voltage\n",
      "\n",
      "Cluster 1\n",
      "reject,size,statistic,condition,probability,alternative,interval,significance,sampling,true,deviation,standard,mean,distribution,confidence,population,proportion,null,hypothesis,sample\n",
      "\n",
      "Cluster 2\n",
      "gradient,gonna,squared,zero,output,like,input,vectors,constant,negative,direction,just,point,kind,component,function,respect,partial,derivative,vector\n",
      "\n",
      "Cluster 3\n",
      "positive,gonna,respect,right,power,slope,squared,equals,just,prime,function,let,times,plus,minus,going,derivative,zero,equal,negative\n",
      "\n",
      "Cluster 4\n",
      "lady,author,semicolon,just,passage,colon,conjunction,questions,photograph,subject,probability,subordinating,dependent,conjunctions,clauses,cost,sentences,independent,sentence,clause\n",
      "\n",
      "Cluster 5\n",
      "power,new,years,court,great,world,rights,really,time,period,amendment,president,american,constitution,united,war,empire,government,states,people\n",
      "\n",
      "Cluster 6\n",
      "enabling,enables,enablers,enabled,enable,en,emma,emulator,emptiness,empress,empowerment,empowering,empowered,empower,employs,employment,employers,emulate,还有直角三角函数,paige\n",
      "\n",
      "Cluster 7\n",
      "points,change,going,increasing,equals,zero,right,theorem,let,closed,differentiable,equal,function,slope,point,value,line,continuous,negative,interval\n",
      "\n",
      "Cluster 8\n",
      "24,length,say,multiplying,divide,power,right,figure,plus,12,just,factor,seven,divided,multiply,let,going,equal,10,times\n",
      "\n",
      "Cluster 9\n",
      "line,way,ll,time,ones,know,think,gonna,point,number,times,say,like,plus,equal,10,just,right,let,going\n",
      "\n",
      "Cluster 10\n",
      "radians,adjacent,square,root,going,complex,let,tangent,right,negative,zero,circle,hypotenuse,equal,degrees,angle,sine,cosine,pi,theta\n",
      "\n",
      "Cluster 11\n",
      "enabled,enable,en,emulator,employee,enacted,emptiness,empowerment,empowering,empowered,empower,employs,employment,employers,employer,employees,empress,enacting,preciou,rosie\n",
      "\n",
      "Cluster 12\n",
      "transportation,just,working,really,excited,going,kind,carrots,soda,like,know,academy,yeah,people,khan,tens,cccccc,color,e5e5e5,font\n",
      "\n",
      "Cluster 13\n",
      "connotation,master,word,metaphor,strawberry,prepositions,present,tense,tenses,past,metaphorical,wizard,verb,literal,david,pleasure,great,talk,sal,thanks\n",
      "\n",
      "Cluster 14\n",
      "seven,positive,infinity,zero,approaching,power,right,just,let,approaches,root,square,going,equal,times,limit,plus,negative,minus,squared\n"
     ]
    }
   ],
   "source": [
    "def get_top_keywords(data, clusters, labels, n_terms):\n",
    "    df = pd.DataFrame(data.todense()).groupby(clusters).mean()\n",
    "    \n",
    "    for i,r in df.iterrows():\n",
    "        print('\\nCluster {}'.format(i))\n",
    "        print(','.join([labels[t] for t in np.argsort(r)[-n_terms:]]))\n",
    "            \n",
    "get_top_keywords(matrix, clusters, tf_idf.get_feature_names(), 20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# let, just, say, 24, 10, like, "
   ]
  }
 ],
 "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": "3.7.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}