feature_selection_aws.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import boto3\n",
    "import pandas as pd\n",
    "from sagemaker import get_execution_role\n",
    "\n",
    "role = get_execution_role()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.feature_selection import SelectKBest\n",
    "from sklearn.feature_selection import mutual_info_classif\n",
    "from matplotlib import pyplot\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "from sklearn.preprocessing import OrdinalEncoder\n",
    "import pandas as pd\n",
    "import matplotlib as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "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>channelId</th>\n",
       "      <th>videoId</th>\n",
       "      <th>viewCount</th>\n",
       "      <th>Friday</th>\n",
       "      <th>Monday</th>\n",
       "      <th>Saturday</th>\n",
       "      <th>Sunday</th>\n",
       "      <th>Thursday</th>\n",
       "      <th>Tuesday</th>\n",
       "      <th>Wednesday</th>\n",
       "      <th>...</th>\n",
       "      <th>titlePRON</th>\n",
       "      <th>titleSYM</th>\n",
       "      <th>titleSCONJ</th>\n",
       "      <th>titleINTJ</th>\n",
       "      <th>titlePART</th>\n",
       "      <th>titleCCONJ</th>\n",
       "      <th>subscriberCount</th>\n",
       "      <th>avgViewCount</th>\n",
       "      <th>humanCount</th>\n",
       "      <th>humanPresence</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>UCjOl2AUblVmg2rA_cRgZkFg</td>\n",
       "      <td>NXX338WY_Lw</td>\n",
       "      <td>184447.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>7610000</td>\n",
       "      <td>2224636.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>UCjOl2AUblVmg2rA_cRgZkFg</td>\n",
       "      <td>dtHcdU2c71Y</td>\n",
       "      <td>217619.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>7610000</td>\n",
       "      <td>2224636.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>UCjOl2AUblVmg2rA_cRgZkFg</td>\n",
       "      <td>vnrtWe-RAzg</td>\n",
       "      <td>437777.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>7610000</td>\n",
       "      <td>2224636.0</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>UCjOl2AUblVmg2rA_cRgZkFg</td>\n",
       "      <td>Ra1F0TsOCPs</td>\n",
       "      <td>191070.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>7610000</td>\n",
       "      <td>2224636.0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>UCjOl2AUblVmg2rA_cRgZkFg</td>\n",
       "      <td>fXysipmTxcQ</td>\n",
       "      <td>572569.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>7610000</td>\n",
       "      <td>2224636.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 61 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                  channelId      videoId  viewCount  Friday  Monday  Saturday  \\\n",
       "0  UCjOl2AUblVmg2rA_cRgZkFg  NXX338WY_Lw   184447.0       1       0         0   \n",
       "1  UCjOl2AUblVmg2rA_cRgZkFg  dtHcdU2c71Y   217619.0       1       0         0   \n",
       "2  UCjOl2AUblVmg2rA_cRgZkFg  vnrtWe-RAzg   437777.0       0       0         0   \n",
       "3  UCjOl2AUblVmg2rA_cRgZkFg  Ra1F0TsOCPs   191070.0       0       0         0   \n",
       "4  UCjOl2AUblVmg2rA_cRgZkFg  fXysipmTxcQ   572569.0       0       0         0   \n",
       "\n",
       "   Sunday  Thursday  Tuesday  Wednesday  ...  titlePRON  titleSYM  titleSCONJ  \\\n",
       "0       0         0        0          0  ...          0         0           0   \n",
       "1       0         0        0          0  ...          0         0           0   \n",
       "2       0         0        0          1  ...          0         0           0   \n",
       "3       0         0        1          0  ...          0         0           0   \n",
       "4       0         0        1          0  ...          0         0           0   \n",
       "\n",
       "   titleINTJ  titlePART  titleCCONJ  subscriberCount  avgViewCount  \\\n",
       "0          0          0           0          7610000     2224636.0   \n",
       "1          0          0           0          7610000     2224636.0   \n",
       "2          0          0           0          7610000     2224636.0   \n",
       "3          0          0           0          7610000     2224636.0   \n",
       "4          0          0           0          7610000     2224636.0   \n",
       "\n",
       "   humanCount  humanPresence  \n",
       "0           0              0  \n",
       "1           0              0  \n",
       "2           2              1  \n",
       "3           1              1  \n",
       "4           0              0  \n",
       "\n",
       "[5 rows x 61 columns]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "featuresdf = pd.read_csv('features_df.csv')\n",
    "featuresdf.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Model 1: ANOVA test for categorical input variables\n",
    "\n",
    "https://machinelearningmastery.com/feature-selection-with-real-and-categorical-data/\n",
    "\n",
    "https://machinelearningmastery.com/feature-selection-with-numerical-input-data/\n",
    "\n",
    "Your results may vary given the stochastic nature of the algorithm or evaluation procedure, or differences in numerical precision. Consider running the example a few times and compare the average outcome.\n",
    "\n",
    "In this case, we can see that some features stand out as perhaps being more relevant than others, with much larger test statistic values."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(405759, 57)\n",
      "Index(['channelId', 'videoId', 'viewCount', 'Friday', 'Monday', 'Saturday',\n",
      "       'Sunday', 'Thursday', 'Tuesday', 'Wednesday', 'Afternoon',\n",
      "       'Early Morning', 'Evening', 'Late Night', 'Morning', 'Night', 'Autumn',\n",
      "       'Spring', 'Summer', 'Winter', 'AUTO & VEHICLES', 'COMEDY', 'EDUCATION',\n",
      "       'ENTERTAINMENT', 'FILM', 'GAMING', 'HOW TO & STYLE', 'MADE FOR KIDS',\n",
      "       'MUSIC', 'NEWS & POLITICS', 'NONPROFIT & ACTIVISM', 'PEOPLE & BLOGS',\n",
      "       'PETS & ANIMALS', 'SCIENCE & TECHNOLOGY', 'SPORTS', 'TRAVEL',\n",
      "       'Negative', 'Neutral', 'Positive', 'titlePROPN', 'titleDET', 'titleADJ',\n",
      "       'titleVERB', 'titleNOUN', 'titleNUM', 'titleAUX', 'titlePUNCT',\n",
      "       'titleADV', 'titleADP', 'titleX', 'titlePRON', 'titleSYM', 'titleSCONJ',\n",
      "       'titleINTJ', 'titlePART', 'titleCCONJ', 'humanPresence'],\n",
      "      dtype='object')\n"
     ]
    }
   ],
   "source": [
    "catdf = featuresdf.drop(columns=['titleLen','subscriberCount','avgViewCount','humanCount'])\n",
    "print(catdf.shape)\n",
    "print(catdf.columns)\n",
    "rows = len(catdf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train (271858, 54) (271858, 1)\n",
      "Test (133901, 54) (133901, 1)\n"
     ]
    }
   ],
   "source": [
    "# Split into input(x) and output(y)\n",
    "\n",
    "#retrieve numpy array & split\n",
    "dataset= catdf.values\n",
    "X = dataset[:,3:]\n",
    "y = dataset[:,2]\n",
    "\n",
    "#format all fields as string\n",
    "X = X.astype(str)\n",
    "y=y.astype('int64')\n",
    "y=y.reshape((rows,1))\n",
    "\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.33, random_state=1)\n",
    "\n",
    "# summarize\n",
    "print('Train', X_train.shape, y_train.shape)\n",
    "print('Test', X_test.shape, y_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/ec2-user/anaconda3/envs/python3/lib/python3.6/site-packages/sklearn/utils/validation.py:755: FutureWarning: Beginning in version 0.22, arrays of bytes/strings will be converted to decimal numbers if dtype='numeric'. It is recommended that you convert the array to a float dtype before using it in scikit-learn, for example by using your_array = your_array.astype(np.float64).\n",
      "  estimator=estimator)\n",
      "/home/ec2-user/anaconda3/envs/python3/lib/python3.6/site-packages/sklearn/utils/validation.py:755: FutureWarning: Beginning in version 0.22, arrays of bytes/strings will be converted to decimal numbers if dtype='numeric'. It is recommended that you convert the array to a float dtype before using it in scikit-learn, for example by using your_array = your_array.astype(np.float64).\n",
      "  estimator=estimator)\n",
      "/home/ec2-user/anaconda3/envs/python3/lib/python3.6/site-packages/sklearn/utils/validation.py:760: DataConversionWarning: A column-vector y was passed when a 1d array was expected. Please change the shape of y to (n_samples, ), for example using ravel().\n",
      "  y = column_or_1d(y, warn=True)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "FEATURE SELECTION DONE\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.063981</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.052212</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.015753</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.009671</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.082198</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          0\n",
       "0  0.063981\n",
       "1  0.052212\n",
       "2  0.015753\n",
       "3  0.009671\n",
       "4  0.082198"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# feature selection\n",
    "def select_features(X_train, y_train, X_test):\n",
    "    # configure to select all features\n",
    "    fs = SelectKBest(score_func=mutual_info_classif, k='all')\n",
    "    # learn relationship from training data\n",
    "    fs.fit(X_train, y_train)\n",
    "    # transform train input data\n",
    "    X_train_fs = fs.transform(X_train)\n",
    "    # transform test input data\n",
    "    X_test_fs = fs.transform(X_test)\n",
    "    return X_train_fs, X_test_fs, fs\n",
    "\n",
    "# feature selection\n",
    "X_train_fs, X_test_fs, fs = select_features(X_train, y_train, X_test)\n",
    "print('FEATURE SELECTION DONE')\n",
    "catscoredf = pd.DataFrame(list(fs.scores_))\n",
    "catscoredf.to_csv('catscore_df.csv', index=False, header=True)\n",
    "catscoredf.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Feature 0: 0.063981\n",
      "Feature 1: 0.052212\n",
      "Feature 2: 0.015753\n",
      "Feature 3: 0.009671\n",
      "Feature 4: 0.082198\n",
      "Feature 5: 0.072116\n",
      "Feature 6: 0.086137\n",
      "Feature 7: 0.198378\n",
      "Feature 8: 0.021491\n",
      "Feature 9: 0.142540\n",
      "Feature 10: 0.091575\n",
      "Feature 11: 0.027779\n",
      "Feature 12: 0.068164\n",
      "Feature 13: 0.157454\n",
      "Feature 14: 0.211570\n",
      "Feature 15: 0.189043\n",
      "Feature 16: 0.180823\n",
      "Feature 17: 0.000541\n",
      "Feature 18: 0.000903\n",
      "Feature 19: 0.004578\n",
      "Feature 20: 0.002140\n",
      "Feature 21: 0.012135\n",
      "Feature 22: 0.000000\n",
      "Feature 23: 0.007958\n",
      "Feature 24: 0.000000\n",
      "Feature 25: 0.000000\n",
      "Feature 26: 2.051160\n",
      "Feature 27: 0.079635\n",
      "Feature 28: 0.000255\n",
      "Feature 29: 0.000164\n",
      "Feature 30: 0.004073\n",
      "Feature 31: 0.182120\n",
      "Feature 32: 0.000438\n",
      "Feature 33: 0.046689\n",
      "Feature 34: 1.596341\n",
      "Feature 35: 0.220304\n",
      "Feature 36: 1.246622\n",
      "Feature 37: 0.005886\n",
      "Feature 38: 0.016149\n",
      "Feature 39: 0.009988\n",
      "Feature 40: 0.043525\n",
      "Feature 41: 0.004214\n",
      "Feature 42: 0.000000\n",
      "Feature 43: 0.003640\n",
      "Feature 44: 0.000352\n",
      "Feature 45: 0.004391\n",
      "Feature 46: 0.000573\n",
      "Feature 47: 0.002832\n",
      "Feature 48: 0.004551\n",
      "Feature 49: 0.000000\n",
      "Feature 50: 0.000000\n",
      "Feature 51: 0.000000\n",
      "Feature 52: 0.000000\n",
      "Feature 53: 2.386547\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": [
    "import numpy as np\n",
    "\n",
    "# what are scores for the features\n",
    "for i in range(len(fs.scores_)):\n",
    "    print('Feature %d: %f' % (i, fs.scores_[i]))\n",
    "\n",
    "# plot the scores\n",
    "pyplot.figure(figsize=(15,10))\n",
    "pyplot.bar([i for i in range(len(fs.scores_))], fs.scores_)\n",
    "pyplot.xticks(np.arange(0,55,5))\n",
    "pyplot.axhline(y=1, color='r', linestyle='-')\n",
    "pyplot.title('Feature Importance Scores of categorical variables')\n",
    "pyplot.xlabel('Feature Index Number')\n",
    "pyplot.ylabel('Feature Importance Score')\n",
    "pyplot.savefig('catbar.png')\n",
    "pyplot.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "26: HOW TO & STYLE\n",
      "34: SPORTS\n",
      "36: Negative\n",
      "53: titleINTJ\n"
     ]
    }
   ],
   "source": [
    "index = []\n",
    "\n",
    "for i in range(len(list(fs.scores_))):\n",
    "    score = list(fs.scores_)[i]\n",
    "    if score>0.5:\n",
    "        print(str(i) + \": \" + catdf.columns[i])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Model 2: Pearson's Correlation test for numerical input variables\n",
    "https://medium.com/analytics-vidhya/feature-selection-techniques-2614b3b7efcd\n",
    "\n",
    "remove variables that are correlated to each other"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(405759, 5)\n",
      "Index(['viewCount', 'titleLen', 'subscriberCount', 'avgViewCount',\n",
      "       'humanCount'],\n",
      "      dtype='object')\n"
     ]
    }
   ],
   "source": [
    "numdf = featuresdf.loc[:,['viewCount','titleLen','subscriberCount','avgViewCount','humanCount']]\n",
    "print(numdf.shape)\n",
    "print(numdf.columns)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "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>viewCount</th>\n",
       "      <th>titleLen</th>\n",
       "      <th>subscriberCount</th>\n",
       "      <th>avgViewCount</th>\n",
       "      <th>humanCount</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>viewCount</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000177</td>\n",
       "      <td>0.091093</td>\n",
       "      <td>0.337469</td>\n",
       "      <td>-0.020841</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>titleLen</th>\n",
       "      <td>0.000177</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.064078</td>\n",
       "      <td>0.022827</td>\n",
       "      <td>-0.016582</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>subscriberCount</th>\n",
       "      <td>0.091093</td>\n",
       "      <td>0.064078</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.254701</td>\n",
       "      <td>-0.062342</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>avgViewCount</th>\n",
       "      <td>0.337469</td>\n",
       "      <td>0.022827</td>\n",
       "      <td>0.254701</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>-0.079744</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>humanCount</th>\n",
       "      <td>-0.020841</td>\n",
       "      <td>-0.016582</td>\n",
       "      <td>-0.062342</td>\n",
       "      <td>-0.079744</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 viewCount  titleLen  subscriberCount  avgViewCount  \\\n",
       "viewCount         1.000000  0.000177         0.091093      0.337469   \n",
       "titleLen          0.000177  1.000000         0.064078      0.022827   \n",
       "subscriberCount   0.091093  0.064078         1.000000      0.254701   \n",
       "avgViewCount      0.337469  0.022827         0.254701      1.000000   \n",
       "humanCount       -0.020841 -0.016582        -0.062342     -0.079744   \n",
       "\n",
       "                 humanCount  \n",
       "viewCount         -0.020841  \n",
       "titleLen          -0.016582  \n",
       "subscriberCount   -0.062342  \n",
       "avgViewCount      -0.079744  \n",
       "humanCount         1.000000  "
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cor = numdf.corr()\n",
    "cor"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "plt.figure(figsize=(10,8))\n",
    "sns.heatmap(cor,annot=True)\n",
    "\n",
    "plt.savefig('heatmap.png')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "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>titleLen</th>\n",
       "      <th>subscriberCount</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>titleLen</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.064078</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>subscriberCount</th>\n",
       "      <td>0.064078</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 titleLen  subscriberCount\n",
       "titleLen         1.000000         0.064078\n",
       "subscriberCount  0.064078         1.000000"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "numdf[['titleLen','subscriberCount']].corr()"
   ]
  },
  {
   "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>titleLen</th>\n",
       "      <th>avgViewCount</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>titleLen</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.022827</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>avgViewCount</th>\n",
       "      <td>0.022827</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              titleLen  avgViewCount\n",
       "titleLen      1.000000      0.022827\n",
       "avgViewCount  0.022827      1.000000"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "numdf[['titleLen','avgViewCount']].corr()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "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>titleLen</th>\n",
       "      <th>humanCount</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>titleLen</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>-0.016582</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>humanCount</th>\n",
       "      <td>-0.016582</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            titleLen  humanCount\n",
       "titleLen    1.000000   -0.016582\n",
       "humanCount -0.016582    1.000000"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "numdf[['titleLen','humanCount']].corr()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "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>subscriberCount</th>\n",
       "      <th>avgViewCount</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>subscriberCount</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.254701</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>avgViewCount</th>\n",
       "      <td>0.254701</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 subscriberCount  avgViewCount\n",
       "subscriberCount         1.000000      0.254701\n",
       "avgViewCount            0.254701      1.000000"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "numdf[['subscriberCount','avgViewCount']].corr()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "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>subscriberCount</th>\n",
       "      <th>humanCount</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>subscriberCount</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>-0.062342</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>humanCount</th>\n",
       "      <td>-0.062342</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 subscriberCount  humanCount\n",
       "subscriberCount         1.000000   -0.062342\n",
       "humanCount             -0.062342    1.000000"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "numdf[['subscriberCount','humanCount']].corr()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "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>avgViewCount</th>\n",
       "      <th>humanCount</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>avgViewCount</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>-0.079744</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>humanCount</th>\n",
       "      <td>-0.079744</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              avgViewCount  humanCount\n",
       "avgViewCount      1.000000   -0.079744\n",
       "humanCount       -0.079744    1.000000"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "numdf[['avgViewCount','humanCount']].corr()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Model 3: ANOVA test for ALL input variables"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(405759, 61)\n",
      "Index(['channelId', 'videoId', 'viewCount', 'Friday', 'Monday', 'Saturday',\n",
      "       'Sunday', 'Thursday', 'Tuesday', 'Wednesday', 'Afternoon',\n",
      "       'Early Morning', 'Evening', 'Late Night', 'Morning', 'Night', 'Autumn',\n",
      "       'Spring', 'Summer', 'Winter', 'AUTO & VEHICLES', 'COMEDY', 'EDUCATION',\n",
      "       'ENTERTAINMENT', 'FILM', 'GAMING', 'HOW TO & STYLE', 'MADE FOR KIDS',\n",
      "       'MUSIC', 'NEWS & POLITICS', 'NONPROFIT & ACTIVISM', 'PEOPLE & BLOGS',\n",
      "       'PETS & ANIMALS', 'SCIENCE & TECHNOLOGY', 'SPORTS', 'TRAVEL',\n",
      "       'titleLen', 'Negative', 'Neutral', 'Positive', 'titlePROPN', 'titleDET',\n",
      "       'titleADJ', 'titleVERB', 'titleNOUN', 'titleNUM', 'titleAUX',\n",
      "       'titlePUNCT', 'titleADV', 'titleADP', 'titleX', 'titlePRON', 'titleSYM',\n",
      "       'titleSCONJ', 'titleINTJ', 'titlePART', 'titleCCONJ', 'subscriberCount',\n",
      "       'avgViewCount', 'humanCount', 'humanPresence'],\n",
      "      dtype='object')\n"
     ]
    }
   ],
   "source": [
    "alldf = featuresdf\n",
    "print(alldf.shape)\n",
    "print(alldf.columns)\n",
    "rows = len(alldf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train (271858, 58) (271858, 1)\n",
      "Test (133901, 58) (133901, 1)\n"
     ]
    }
   ],
   "source": [
    "# Split into input(x) and output(y)\n",
    "\n",
    "#retrieve numpy array & split\n",
    "dataset= alldf.values\n",
    "X = dataset[:,3:]\n",
    "y = dataset[:,2]\n",
    "\n",
    "#format all fields as string\n",
    "X = X.astype(str)\n",
    "y=y.astype('int64')\n",
    "y=y.reshape((rows,1))\n",
    "\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.33, random_state=1)\n",
    "\n",
    "# summarize\n",
    "print('Train', X_train.shape, y_train.shape)\n",
    "print('Test', X_test.shape, y_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/ec2-user/anaconda3/envs/python3/lib/python3.6/site-packages/sklearn/utils/validation.py:755: FutureWarning: Beginning in version 0.22, arrays of bytes/strings will be converted to decimal numbers if dtype='numeric'. It is recommended that you convert the array to a float dtype before using it in scikit-learn, for example by using your_array = your_array.astype(np.float64).\n",
      "  estimator=estimator)\n",
      "/home/ec2-user/anaconda3/envs/python3/lib/python3.6/site-packages/sklearn/utils/validation.py:755: FutureWarning: Beginning in version 0.22, arrays of bytes/strings will be converted to decimal numbers if dtype='numeric'. It is recommended that you convert the array to a float dtype before using it in scikit-learn, for example by using your_array = your_array.astype(np.float64).\n",
      "  estimator=estimator)\n",
      "/home/ec2-user/anaconda3/envs/python3/lib/python3.6/site-packages/sklearn/utils/validation.py:760: DataConversionWarning: A column-vector y was passed when a 1d array was expected. Please change the shape of y to (n_samples, ), for example using ravel().\n",
      "  y = column_or_1d(y, warn=True)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "FEATURE SELECTION DONE\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.061123</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.053195</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.015314</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.007945</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.083718</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          0\n",
       "0  0.061123\n",
       "1  0.053195\n",
       "2  0.015314\n",
       "3  0.007945\n",
       "4  0.083718"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# feature selection\n",
    "def select_features(X_train, y_train, X_test):\n",
    "    # configure to select all features\n",
    "    fs = SelectKBest(score_func=mutual_info_classif, k='all')\n",
    "    # learn relationship from training data\n",
    "    fs.fit(X_train, y_train)\n",
    "    # transform train input data\n",
    "    X_train_fs = fs.transform(X_train)\n",
    "    # transform test input data\n",
    "    X_test_fs = fs.transform(X_test)\n",
    "    return X_train_fs, X_test_fs, fs\n",
    "\n",
    "# feature selection\n",
    "X_train_fs, X_test_fs, fs = select_features(X_train, y_train, X_test)\n",
    "print('FEATURE SELECTION DONE')\n",
    "allscoredf = pd.DataFrame(list(fs.scores_))\n",
    "allscoredf.to_csv('allscore_df.csv', index=False, header=True)\n",
    "allscoredf.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Feature 0: 0.061123\n",
      "Feature 1: 0.053195\n",
      "Feature 2: 0.015314\n",
      "Feature 3: 0.007945\n",
      "Feature 4: 0.083718\n",
      "Feature 5: 0.073411\n",
      "Feature 6: 0.082686\n",
      "Feature 7: 0.193197\n",
      "Feature 8: 0.022390\n",
      "Feature 9: 0.142687\n",
      "Feature 10: 0.093379\n",
      "Feature 11: 0.027605\n",
      "Feature 12: 0.062872\n",
      "Feature 13: 0.154480\n",
      "Feature 14: 0.206763\n",
      "Feature 15: 0.187820\n",
      "Feature 16: 0.179872\n",
      "Feature 17: 0.000397\n",
      "Feature 18: 0.000000\n",
      "Feature 19: 0.001589\n",
      "Feature 20: 0.002956\n",
      "Feature 21: 0.015832\n",
      "Feature 22: 0.000000\n",
      "Feature 23: 0.003377\n",
      "Feature 24: 0.000000\n",
      "Feature 25: 0.000000\n",
      "Feature 26: 2.041418\n",
      "Feature 27: 0.073173\n",
      "Feature 28: 0.000000\n",
      "Feature 29: 0.001595\n",
      "Feature 30: 0.000722\n",
      "Feature 31: 0.181007\n",
      "Feature 32: 0.000000\n",
      "Feature 33: 0.156247\n",
      "Feature 34: 0.042901\n",
      "Feature 35: 1.582291\n",
      "Feature 36: 0.218732\n",
      "Feature 37: 1.244789\n",
      "Feature 38: 0.008676\n",
      "Feature 39: 0.017713\n",
      "Feature 40: 0.004064\n",
      "Feature 41: 0.046059\n",
      "Feature 42: 0.001063\n",
      "Feature 43: 0.000000\n",
      "Feature 44: 0.000000\n",
      "Feature 45: 0.002421\n",
      "Feature 46: 0.002807\n",
      "Feature 47: 0.000000\n",
      "Feature 48: 0.000000\n",
      "Feature 49: 0.003904\n",
      "Feature 50: 0.002812\n",
      "Feature 51: 0.000000\n",
      "Feature 52: 0.001211\n",
      "Feature 53: 0.000000\n",
      "Feature 54: 0.396825\n",
      "Feature 55: 0.435447\n",
      "Feature 56: 0.624370\n",
      "Feature 57: 2.389420\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": [
    "import numpy as np\n",
    "\n",
    "# what are scores for the features\n",
    "for i in range(len(fs.scores_)):\n",
    "    print('Feature %d: %f' % (i, fs.scores_[i]))\n",
    "    \n",
    "# plot the scores\n",
    "pyplot.figure(figsize=(15,10))\n",
    "pyplot.bar([i for i in range(len(fs.scores_))], fs.scores_)\n",
    "pyplot.xticks(np.arange(0,60,5))\n",
    "pyplot.axhline(y=1, color='r', linestyle='-')\n",
    "pyplot.title('Feature Importance Scores of all variables')\n",
    "pyplot.xlabel('Feature Index Number')\n",
    "pyplot.ylabel('Feature Importance Score')\n",
    "pyplot.savefig('allbar.png')\n",
    "pyplot.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "26: HOW TO & STYLE\n",
      "35: TRAVEL\n",
      "37: Negative\n",
      "56: titleCCONJ\n",
      "57: subscriberCount\n"
     ]
    }
   ],
   "source": [
    "index = []\n",
    "\n",
    "for i in range(len(list(fs.scores_))):\n",
    "    score = list(fs.scores_)[i]\n",
    "    if score>1:\n",
    "        print(str(i) + \": \" + alldf.columns[i])"
   ]
  }
 ],
 "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": "3.8.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}