Rating_Analysis.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "ymsmTx6J01iD",
    "outputId": "5f4c014a-c45b-47d5-ca90-1fdc77db4257"
   },
   "outputs": [],
   "source": [
    "import numpy as np \n",
    "import pandas as pd\n",
    "import heapq\n",
    "import warnings\n",
    "from pandas.core.common import SettingWithCopyWarning\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from collections import Counter\n",
    "import sklearn\n",
    "from sklearn.cluster import KMeans\n",
    "import random\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn import metrics\n",
    "from sklearn.feature_selection import chi2\n",
    "from sklearn.ensemble import ExtraTreesClassifier\n",
    "from sklearn.preprocessing import StandardScaler, MinMaxScaler, MaxAbsScaler, QuantileTransformer\n",
    "from scipy import spatial\n",
    "\n",
    "\n",
    "import tensorflow as tf\n",
    "import tensorflow_hub as hub\n",
    "# module_url = \"https://tfhub.dev/google/universal-sentence-encoder/4\"\n",
    "# model = hub.load(module_url)\n",
    "# def embed(input):\n",
    "#   return model(input)\n",
    "\n",
    "# nltk.download('punkt')\n",
    "# nltk.download('stopwords')\n",
    "\n",
    "import sklearn\n",
    "from sklearn.ensemble import RandomForestRegressor\n",
    "from sklearn import metrics\n",
    "from sklearn.model_selection import cross_val_predict\n",
    "from sklearn.model_selection import cross_val_score\n",
    "from sklearn.model_selection import cross_validate\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.model_selection import RandomizedSearchCV\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn.linear_model import LinearRegression\n",
    "from sklearn.preprocessing import RobustScaler\n",
    "from pprint import pprint\n",
    "\n",
    "warnings.simplefilter(action=\"ignore\", category=SettingWithCopyWarning)\n",
    "sns.set_style(\"darkgrid\")\n",
    "# from google.colab import drive\n",
    "# drive.mount('/content/drive')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "iQzOi2IBiVqu"
   },
   "source": [
    "## Loading datasets:\n",
    "In order to minimize cultural differences, only movies that **USA** has been on of their production countries have been chosen. Also, all movies without a box office record (**usa gross** or domestic gross) have been droped.\n",
    "Lastly, only movies **after 1990** have been considered for this task"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "id": "uMBxrI5OiapM"
   },
   "outputs": [],
   "source": [
    "movie_df = pd.read_csv('imdb_movies.csv', low_memory=False)\n",
    "rating_df = pd.read_csv('imdb_ratings.csv')\n",
    "movie_df = movie_df[movie_df['country'].str.contains('USA')]\n",
    "movie_df = movie_df[movie_df.u_gross_money.notnull()]\n",
    "movie_df = movie_df[movie_df['title_year'] >= 1990]\n",
    "movie_df.set_index('movie_id', inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "6yzeYmmfDMZQ",
    "outputId": "ab0795e3-3e3c-4e79-8b91-d8655a3f4852"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "10231"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(movie_df)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "UckBr4IPjHQF"
   },
   "source": [
    "## Feature engineering:"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "IQLHOXlkjMXF"
   },
   "source": [
    "#### Checking available columns and removing unnecessary ones:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "l7de193zjItd",
    "outputId": "a3378dd8-c2f0-4aa8-a046-36c43a8ce3ed"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['name', 'org_name', 'date', 'title_year', 'point', 'point_volume',\n",
       "       'metascore', 'user_reviews', 'critic_reviews', 'director', 'writer',\n",
       "       'story_line', 'cast', 'genres', 'country', 'language', 'budget',\n",
       "       'world_gross', 'usa_gross', 'runtime', 'production_companies',\n",
       "       'dollar_budget', 'w_gross_money', 'u_gross_money', 'inflation_coeff',\n",
       "       'casts_id', 'BlogPage', 'CompPage', 'HomePage', 'release_month',\n",
       "       'release_day', 'keywords'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "movie_df.columns"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "M2Ok6-CAjTl7"
   },
   "source": [
    "We will not be needing **name**, **original name**, **date of release** (since we have year, month and day of release as separate columns), **point_volume**, **metascore**, **budget** (we have dollar_budget column). **World gross** , **usa gross** and their corresponding **w_gross_money** and **u_gross_money** should also be dropped since there's no real value in a algorithm that needs to know movie gross in order to predict it's rating.\n",
    "\n",
    "**We will use keywords and storyline for Text similary predictor!**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 582
    },
    "id": "E69R1xFgjY_O",
    "outputId": "0200cd89-a74b-4e57-b941-747aa2e77147"
   },
   "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>title_year</th>\n",
       "      <th>point</th>\n",
       "      <th>director</th>\n",
       "      <th>writer</th>\n",
       "      <th>story_line</th>\n",
       "      <th>cast</th>\n",
       "      <th>genres</th>\n",
       "      <th>country</th>\n",
       "      <th>language</th>\n",
       "      <th>runtime</th>\n",
       "      <th>production_companies</th>\n",
       "      <th>dollar_budget</th>\n",
       "      <th>inflation_coeff</th>\n",
       "      <th>BlogPage</th>\n",
       "      <th>CompPage</th>\n",
       "      <th>HomePage</th>\n",
       "      <th>release_month</th>\n",
       "      <th>release_day</th>\n",
       "      <th>keywords</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>movie_id</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>tt0035423</th>\n",
       "      <td>2001.0</td>\n",
       "      <td>6.4</td>\n",
       "      <td>James Mangold</td>\n",
       "      <td>Steven Rogers,James Mangold</td>\n",
       "      <td>Kate and her actor brother live in N.Y. in the...</td>\n",
       "      <td>Meg Ryan,Liev Schreiber,Natasha Lyonne,Paxton ...</td>\n",
       "      <td>Comedy,Fantasy,Romance</td>\n",
       "      <td>USA</td>\n",
       "      <td>English</td>\n",
       "      <td>118.0</td>\n",
       "      <td>Konrad Pictures,Miramax</td>\n",
       "      <td>48000000.0</td>\n",
       "      <td>1.470013</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>12.0</td>\n",
       "      <td>25.0</td>\n",
       "      <td>time-travel,brooklyn-bridge,bridge,time-travel...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>tt0081145</th>\n",
       "      <td>1993.0</td>\n",
       "      <td>5.4</td>\n",
       "      <td>Dan Curtis</td>\n",
       "      <td>Stanley Cohen,Richard Tannenbaum</td>\n",
       "      <td>Two ex-cons, Harry (Aiello) and Roy (Pantolian...</td>\n",
       "      <td>Danny Aiello,Joe Pantoliano,David Dukes,Rick A...</td>\n",
       "      <td>Comedy,Crime,Drama,Family</td>\n",
       "      <td>USA</td>\n",
       "      <td>English</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Dan Curtis Productions</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1.800164</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>10.0</td>\n",
       "      <td>22.0</td>\n",
       "      <td>kidnapper-kidnapped-relationship,family-relati...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>tt0097115</th>\n",
       "      <td>1990.0</td>\n",
       "      <td>5.7</td>\n",
       "      <td>Christopher Leitch</td>\n",
       "      <td>Fred Brogger,Mark Brogger</td>\n",
       "      <td>In this version of \"Heidi,\" a young Swiss girl...</td>\n",
       "      <td>Juliette Caton,Nicola Stapleton,Kathryn Ludlow...</td>\n",
       "      <td>Adventure,Drama,Family,War</td>\n",
       "      <td>USA,France</td>\n",
       "      <td>English</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Epic Productions,Stone Group Pictures</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1.992922</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>16.0</td>\n",
       "      <td>record-player,mountain-climbing,carriage,homes...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>tt0098375</th>\n",
       "      <td>1990.0</td>\n",
       "      <td>4.8</td>\n",
       "      <td>Tobe Hooper</td>\n",
       "      <td>Tobe Hooper,Tobe Hooper</td>\n",
       "      <td>A young man finds out that his parents had bee...</td>\n",
       "      <td>Brad Dourif,Jon Cypher,Melinda Dillon,Tegan We...</td>\n",
       "      <td>Horror,Sci-Fi,Thriller</td>\n",
       "      <td>USA</td>\n",
       "      <td>English</td>\n",
       "      <td>97.0</td>\n",
       "      <td>Black Owl Productions,Project Samson,Skyland S...</td>\n",
       "      <td>5500000.0</td>\n",
       "      <td>1.992922</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>23.0</td>\n",
       "      <td>experiment-gone-wrong,human-experiment,science...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>tt0098966</th>\n",
       "      <td>1990.0</td>\n",
       "      <td>5.5</td>\n",
       "      <td>Emile Ardolino</td>\n",
       "      <td>Coline Serreau,Sara Parriott</td>\n",
       "      <td>Five happy years after baby Mary's arrival in</td>\n",
       "      <td>Tom Selleck,Ted Danson,Robin Weisman,Sheila Ha...</td>\n",
       "      <td>Comedy,Drama,Family,Romance</td>\n",
       "      <td>USA</td>\n",
       "      <td>English</td>\n",
       "      <td>104.0</td>\n",
       "      <td>Touchstone Pictures,Interscope Communications,...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1.992922</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>11.0</td>\n",
       "      <td>21.0</td>\n",
       "      <td>sequel-to-remake,second-part,sequel,girl,numbe...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           title_year  point            director  \\\n",
       "movie_id                                           \n",
       "tt0035423      2001.0    6.4       James Mangold   \n",
       "tt0081145      1993.0    5.4          Dan Curtis   \n",
       "tt0097115      1990.0    5.7  Christopher Leitch   \n",
       "tt0098375      1990.0    4.8         Tobe Hooper   \n",
       "tt0098966      1990.0    5.5      Emile Ardolino   \n",
       "\n",
       "                                     writer  \\\n",
       "movie_id                                      \n",
       "tt0035423       Steven Rogers,James Mangold   \n",
       "tt0081145  Stanley Cohen,Richard Tannenbaum   \n",
       "tt0097115         Fred Brogger,Mark Brogger   \n",
       "tt0098375           Tobe Hooper,Tobe Hooper   \n",
       "tt0098966      Coline Serreau,Sara Parriott   \n",
       "\n",
       "                                                  story_line  \\\n",
       "movie_id                                                       \n",
       "tt0035423  Kate and her actor brother live in N.Y. in the...   \n",
       "tt0081145  Two ex-cons, Harry (Aiello) and Roy (Pantolian...   \n",
       "tt0097115  In this version of \"Heidi,\" a young Swiss girl...   \n",
       "tt0098375  A young man finds out that his parents had bee...   \n",
       "tt0098966     Five happy years after baby Mary's arrival in    \n",
       "\n",
       "                                                        cast  \\\n",
       "movie_id                                                       \n",
       "tt0035423  Meg Ryan,Liev Schreiber,Natasha Lyonne,Paxton ...   \n",
       "tt0081145  Danny Aiello,Joe Pantoliano,David Dukes,Rick A...   \n",
       "tt0097115  Juliette Caton,Nicola Stapleton,Kathryn Ludlow...   \n",
       "tt0098375  Brad Dourif,Jon Cypher,Melinda Dillon,Tegan We...   \n",
       "tt0098966  Tom Selleck,Ted Danson,Robin Weisman,Sheila Ha...   \n",
       "\n",
       "                                genres     country language  runtime  \\\n",
       "movie_id                                                               \n",
       "tt0035423       Comedy,Fantasy,Romance         USA  English    118.0   \n",
       "tt0081145    Comedy,Crime,Drama,Family         USA  English      NaN   \n",
       "tt0097115   Adventure,Drama,Family,War  USA,France  English      NaN   \n",
       "tt0098375       Horror,Sci-Fi,Thriller         USA  English     97.0   \n",
       "tt0098966  Comedy,Drama,Family,Romance         USA  English    104.0   \n",
       "\n",
       "                                        production_companies  dollar_budget  \\\n",
       "movie_id                                                                      \n",
       "tt0035423                            Konrad Pictures,Miramax     48000000.0   \n",
       "tt0081145                             Dan Curtis Productions            NaN   \n",
       "tt0097115              Epic Productions,Stone Group Pictures            NaN   \n",
       "tt0098375  Black Owl Productions,Project Samson,Skyland S...      5500000.0   \n",
       "tt0098966  Touchstone Pictures,Interscope Communications,...            NaN   \n",
       "\n",
       "           inflation_coeff  BlogPage  CompPage  HomePage  release_month  \\\n",
       "movie_id                                                                  \n",
       "tt0035423         1.470013         0         0         0           12.0   \n",
       "tt0081145         1.800164         0         0         0           10.0   \n",
       "tt0097115         1.992922         0         0         0            2.0   \n",
       "tt0098375         1.992922         0         0         0            2.0   \n",
       "tt0098966         1.992922         0         0         0           11.0   \n",
       "\n",
       "           release_day                                           keywords  \n",
       "movie_id                                                                   \n",
       "tt0035423         25.0  time-travel,brooklyn-bridge,bridge,time-travel...  \n",
       "tt0081145         22.0  kidnapper-kidnapped-relationship,family-relati...  \n",
       "tt0097115         16.0  record-player,mountain-climbing,carriage,homes...  \n",
       "tt0098375         23.0  experiment-gone-wrong,human-experiment,science...  \n",
       "tt0098966         21.0  sequel-to-remake,second-part,sequel,girl,numbe...  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "movie_df.drop(['name', 'org_name', 'budget', 'world_gross', 'usa_gross', 'date', 'user_reviews', 'critic_reviews',\n",
    "              'w_gross_money', 'u_gross_money', 'metascore', 'point_volume', 'casts_id'], axis = 1, inplace=True)\n",
    "movie_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "DaKMgfu5j-2w"
   },
   "source": [
    "#### Adding inflation to budget and scaling:"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "8Y3a4YfxkS0x"
   },
   "source": [
    "Also, we should multiply dollar_budget and inflation_coeff and save the result as dolla_budget. If we ignore this part, we are ignoring the fact that 1 dollar in the 20's is equal to about 14 dollars now. Next, we will normalize budget values with a simple standard scaler. This action will be of aid in predicting ratings and reducing noise."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "id": "IKD0UnItkB5j"
   },
   "outputs": [],
   "source": [
    "standard_scaler = StandardScaler()\n",
    "min_max_scaler = MinMaxScaler()\n",
    "movie_df['dollar_budget'] = movie_df['dollar_budget'] * movie_df['inflation_coeff']\n",
    "movie_df['dollar_budget'] = min_max_scaler.fit_transform(movie_df[['dollar_budget']])\n",
    "movie_df.drop(['inflation_coeff'], axis = 1, inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "lBRomofjjgNY"
   },
   "source": [
    "#### Looking for NaN values:\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "0_lIRoQbjigj",
    "outputId": "dc18d7a8-4b2f-462f-c812-0e722d4a094c"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['point', 'runtime', 'dollar_budget'], dtype='object')"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "movie_df.loc[:, movie_df.isna().any()].columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "id": "q-so0a5rjpj7"
   },
   "outputs": [],
   "source": [
    "## drop movies withiout point or dollar_budget values:\n",
    "movie_df.dropna(subset= ['point', 'dollar_budget'], how='any', inplace=True)\n",
    "\n",
    "## fillna runtime based on main genre and mean of all other movies:\n",
    "drama_runtime_mean = movie_df[movie_df['genres'] == 'Drama'].runtime.mean()\n",
    "doc_runtime_mean = movie_df[movie_df['genres'] == 'Documentary'].runtime.mean()\n",
    "comedy_runtime_mean = movie_df[movie_df['genres'] == 'Comedy'].runtime.mean()\n",
    "horror_runtime_mean = movie_df[movie_df['genres'] == 'Horror'].runtime.mean()\n",
    "movie_df['runtime'] = np.where(movie_df['genres'] == 'Drama', drama_runtime_mean, movie_df['runtime'])\n",
    "movie_df['runtime'] = np.where(movie_df['genres'] == 'Documentary', doc_runtime_mean, movie_df['runtime'])\n",
    "movie_df['runtime'] = np.where(movie_df['genres'] == 'Comedy', comedy_runtime_mean, movie_df['runtime'])\n",
    "movie_df['runtime'] = np.where(movie_df['genres'] == 'Horror', horror_runtime_mean, movie_df['runtime'])\n",
    "movie_df['runtime'].fillna((movie_df['runtime'].mean()), inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Hef3x-1ekMzT"
   },
   "source": [
    "#### Label Encoding utility function:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "id": "06jNTeAFkOXX"
   },
   "outputs": [],
   "source": [
    "def get_best_columns(base_col, count=10):\n",
    "    all_contents = list()\n",
    "    for index, row in movie_df.iterrows():\n",
    "        content = row[base_col]\n",
    "        chunks = content.split(',')\n",
    "        for i in chunks:\n",
    "          all_contents.append(i)\n",
    "    c = Counter(all_contents)\n",
    "    most_common = [key for key, val in c.most_common(count)]\n",
    "    return most_common"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "yo9fWx79kYuq"
   },
   "source": [
    "#### Label Encoding Genres:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "id": "VwxHaaZ3kfGY"
   },
   "outputs": [],
   "source": [
    "genres_labels = movie_df['genres'].str.get_dummies(sep=',')\n",
    "most_common = get_best_columns('genres', 5)\n",
    "genres_labels = genres_labels[most_common]\n",
    "movie_df = pd.merge(movie_df, genres_labels, left_index=True, right_index=True)\n",
    "movie_df.drop(['genres'], axis = 1, inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "p0W8OAX3ki7E"
   },
   "source": [
    "#### Label Encoding Countries:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "id": "hj21EtZdkk1j"
   },
   "outputs": [],
   "source": [
    "countries_labels = movie_df['country'].str.get_dummies(sep=',')\n",
    "most_common = get_best_columns('country')\n",
    "countries_labels = countries_labels[most_common]\n",
    "movie_df = pd.merge(movie_df, countries_labels, left_index=True, right_index=True)\n",
    "movie_df.drop(['country'], axis = 1, inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "CI_HEBwLkstp"
   },
   "source": [
    "#### Label Encoding Languages:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "6sMe8P3YkpEE",
    "outputId": "2a6b7d0e-b453-4c97-e71f-beb3f8612969"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "English    0.995406\n",
      "Spanish    0.047984\n",
      "French     0.045772\n",
      "German     0.028246\n",
      "Italian    0.021269\n",
      "dtype: float64\n"
     ]
    }
   ],
   "source": [
    "language_labels = movie_df['language'].str.get_dummies(sep=',')\n",
    "most_common = get_best_columns('language')\n",
    "language_labels = language_labels[most_common]\n",
    "movie_df = pd.merge(movie_df, language_labels, left_index=True, right_index=True)\n",
    "print(language_labels.mean().sort_values(ascending=False)[0:5])\n",
    "movie_df.drop(['language'], axis = 1, inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "punri2lqkwuS"
   },
   "source": [
    "#### Label Encoding Production Companies:\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "id": "U49Qlh1FkxPP"
   },
   "outputs": [],
   "source": [
    "companies_list = get_best_columns('production_companies', 6)\n",
    "if ('company_not_provided' in companies_list):\n",
    "  companies_list.remove('company_not_provided')\n",
    "for i in companies_list:\n",
    "    movie_df[i] = 0\n",
    "for index, row in movie_df.iterrows():\n",
    "    for i in companies_list:\n",
    "        if(i in str(row['production_companies'])):\n",
    "            movie_df.loc[index, i] = 1\n",
    "\n",
    "movie_df.drop(['production_companies'], axis = 1, inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "kij4JVL2k2jd"
   },
   "source": [
    "#### Label Encoding Casts:\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "D-ajPVvOk7o3"
   },
   "source": [
    "There are thousands of casts out there. We will use https://stephenfollows.com/which-actors-most-frequently-appear-in-excellent-awful-movies/ in order to find those actors that most frequently appear in highly rated movies. We will use top actors and see how many of these legends each movie has as it's cast."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "id": "Ns7qboGKk8Hb"
   },
   "outputs": [],
   "source": [
    "god_actors = ['Carey Mulligan', 'Tilda Swinton', 'Tom Hardy', 'Andy Serkis',\n",
    "            'Ralph Fiennes', 'Tom Hanks', 'Domhnall Gleeson', 'George Clooney',\n",
    "            'Michael Stuhlbarg', 'Ryan Gosling', 'Emma Watson', 'Leonardo DiCaprio',\n",
    "            'Benedict Cumberbatch', 'Samuel L. Jackson', 'Mark Falvo', 'Morgan Freeman',\n",
    "            'Robert De Niro', 'Ian McKellen', 'Philip Seymour Hoffman', 'Alan Tudyk',\n",
    "              'Al Pacino', 'Christian Bale', 'Denzel Washington', 'Anthony Hopkins',\n",
    "              'Brad Pitt', 'Jack Nicholson', 'Dustin Hoffman', 'Gary Oldman', 'Joaquin Phoenix',\n",
    "              'Matt Damon', 'Will Smith', 'Harrison Ford', 'Tom Cruise', 'Johnny Depp',\n",
    "              'Meryl Streep', 'Daniel Day-Lewis']\n",
    "\n",
    "movie_df['god_actors'] = 0\n",
    "\n",
    "for index,row in movie_df.iterrows():\n",
    "    for god in god_actors:\n",
    "        if(god in str(row['cast'])):\n",
    "            movie_df.loc[index, 'god_actors'] += 1\n",
    "\n",
    "\n",
    "movie_df.drop(['cast'], axis = 1, inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "H4yOZhyek-m1"
   },
   "source": [
    "#### Lastly, Writers and directors"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "8U9g0JxklD_P"
   },
   "source": [
    "Let's use the List of highest-grossing (and highest rated) film directors(or writers). This list is available at https://en.wikipedia.org/wiki/List_of_highest-grossing_film_directors and https://www.imdb.com/list/ls076197867/. We will use top directors and see if each movie does have at least one of these legends as either director or writer."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "id": "5hu89_EzlHz8"
   },
   "outputs": [],
   "source": [
    "gods_list = ['Steven Spielberg', 'Anthony Russo', 'James Cameron', 'Michael Bay',\n",
    "            'Peter Jackson', 'David Yates', 'Christopher Nolan', 'J.J. Abrams',\n",
    "            'Tim Burton', 'Jon Favreau', 'Ron Howard', 'Clint Eastwood', 'Martin Scorsese',\n",
    "            'Alfonso Cuarón', 'Quentin Tarantino', 'Woody Allen', 'David Fincher',\n",
    "             'Francis Ford Coppola', 'Ridley Scott', 'David Lynch', 'Guy Ritchie', \n",
    "             'Ang Lee', 'Wes Anderson', 'Robert Zemeckis', 'Alfred Hitchcock',\n",
    "             'George Lucas', 'Denis Villeneuve', 'Paul Thomas Anderson', 'Terrence Malick',\n",
    "             'Ethan Coen', 'Charlie Kaufman', 'Aaron Sorkin', 'David Mamet', 'Christopher McQuarrie',\n",
    "             'Frank Darabont', 'Richard Linklater', 'Spike Lee', 'Rob Reiner', 'James L. Brooks',\n",
    "             'Oliver Stone', 'Terry Gilliam', 'Michael Mann']\n",
    "\n",
    "movie_df['god_wridir'] = 0\n",
    "\n",
    "for index,row in movie_df.iterrows():\n",
    "    for god in gods_list:\n",
    "        if(god in str(row['director']) or god in str(row['writer'])):\n",
    "            movie_df.loc[index, 'god_wridir'] = 1\n",
    "            \n",
    "movie_df.drop(['writer', 'director'], axis = 1, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[nltk_data] Downloading package punkt to\n",
      "[nltk_data]     C:\\Users\\hoomo\\AppData\\Roaming\\nltk_data...\n",
      "[nltk_data]   Package punkt is already up-to-date!\n",
      "[nltk_data] Downloading package stopwords to\n",
      "[nltk_data]     C:\\Users\\hoomo\\AppData\\Roaming\\nltk_data...\n",
      "[nltk_data]   Package stopwords is already up-to-date!\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import nltk\n",
    "from wordcloud import WordCloud, STOPWORDS\n",
    "from nltk.corpus import stopwords\n",
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "from nltk.tokenize import word_tokenize\n",
    "import matplotlib.pyplot as plt\n",
    "nltk.download('punkt')\n",
    "nltk.download('stopwords')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.subplots(figsize=(12,12))\n",
    "stop_words = set(stopwords.words('english'))\n",
    "stop_words.update(',',';','!','?','.','(',')','$','#','+',':','...',' ','')\n",
    "words = movie_df['keywords'].dropna().apply(nltk.word_tokenize)\n",
    "word=[]\n",
    "for i in words:\n",
    "      word.extend(i)\n",
    "word=pd.Series(word)\n",
    "word=([i for i in word.str.lower() if i not in stop_words])\n",
    "wc = WordCloud(background_color=\"black\", max_words=50, stopwords=STOPWORDS, max_font_size= 60,width=1000,height=1000)\n",
    "wc.generate(\" \".join(word))\n",
    "plt.imshow(wc)\n",
    "plt.axis('off')\n",
    "fig=plt.gcf()\n",
    "fig.set_size_inches(8,8)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['murder',\n",
       " 'death',\n",
       " 'husband-wife-relationship',\n",
       " 'violence',\n",
       " 'flashback',\n",
       " 'father-son-relationship',\n",
       " 'blood',\n",
       " 'title-spoken-by-character',\n",
       " 'bare-chested-male',\n",
       " 'friendship']"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from collections import Counter\n",
    "c = Counter(word)\n",
    "n = 10 \n",
    "most_common = [key for key, val in c.most_common(n)]\n",
    "most_common"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "for j in most_common: \n",
    "    movie_df[j] = 0\n",
    "    \n",
    "for index, row in movie_df.iterrows():\n",
    "    for j in most_common:\n",
    "        if (j in row['keywords']):\n",
    "           movie_df.loc[index, j] = 1 "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "2MMWd5i6lMed"
   },
   "source": [
    "#### Last checks:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 287
    },
    "id": "R-j8ivAhlLRg",
    "outputId": "b575866e-3f36-4001-bbba-82f2662867d2"
   },
   "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>title_year</th>\n",
       "      <th>point</th>\n",
       "      <th>runtime</th>\n",
       "      <th>dollar_budget</th>\n",
       "      <th>BlogPage</th>\n",
       "      <th>CompPage</th>\n",
       "      <th>HomePage</th>\n",
       "      <th>release_month</th>\n",
       "      <th>release_day</th>\n",
       "      <th>Drama</th>\n",
       "      <th>...</th>\n",
       "      <th>murder</th>\n",
       "      <th>death</th>\n",
       "      <th>husband-wife-relationship</th>\n",
       "      <th>violence</th>\n",
       "      <th>flashback</th>\n",
       "      <th>father-son-relationship</th>\n",
       "      <th>blood</th>\n",
       "      <th>title-spoken-by-character</th>\n",
       "      <th>bare-chested-male</th>\n",
       "      <th>friendship</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>movie_id</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>tt0035423</th>\n",
       "      <td>2001.0</td>\n",
       "      <td>6.4</td>\n",
       "      <td>118.0</td>\n",
       "      <td>0.187416</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>12.0</td>\n",
       "      <td>25.0</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>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>tt0098375</th>\n",
       "      <td>1990.0</td>\n",
       "      <td>4.8</td>\n",
       "      <td>97.0</td>\n",
       "      <td>0.029114</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>23.0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>tt0098987</th>\n",
       "      <td>1990.0</td>\n",
       "      <td>6.5</td>\n",
       "      <td>104.0</td>\n",
       "      <td>0.105868</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>7.0</td>\n",
       "      <td>11.0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>1</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>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>tt0099005</th>\n",
       "      <td>1990.0</td>\n",
       "      <td>5.8</td>\n",
       "      <td>113.0</td>\n",
       "      <td>0.185269</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>8.0</td>\n",
       "      <td>10.0</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>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>tt0099012</th>\n",
       "      <td>1990.0</td>\n",
       "      <td>6.6</td>\n",
       "      <td>106.0</td>\n",
       "      <td>0.063521</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>10.0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 52 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "           title_year  point  runtime  dollar_budget  BlogPage  CompPage  \\\n",
       "movie_id                                                                   \n",
       "tt0035423      2001.0    6.4    118.0       0.187416         0         0   \n",
       "tt0098375      1990.0    4.8     97.0       0.029114         0         0   \n",
       "tt0098987      1990.0    6.5    104.0       0.105868         0         0   \n",
       "tt0099005      1990.0    5.8    113.0       0.185269         0         0   \n",
       "tt0099012      1990.0    6.6    106.0       0.063521         0         0   \n",
       "\n",
       "           HomePage  release_month  release_day  Drama  ...  murder  death  \\\n",
       "movie_id                                                ...                  \n",
       "tt0035423         0           12.0         25.0      0  ...       0      0   \n",
       "tt0098375         0            2.0         23.0      0  ...       1      1   \n",
       "tt0098987         0            7.0         11.0      0  ...       1      1   \n",
       "tt0099005         0            8.0         10.0      0  ...       0      0   \n",
       "tt0099012         0            1.0         10.0      0  ...       0      0   \n",
       "\n",
       "           husband-wife-relationship  violence  flashback  \\\n",
       "movie_id                                                    \n",
       "tt0035423                          0         0          0   \n",
       "tt0098375                          0         1          1   \n",
       "tt0098987                          0         0          0   \n",
       "tt0099005                          0         0          0   \n",
       "tt0099012                          1         0          1   \n",
       "\n",
       "           father-son-relationship  blood  title-spoken-by-character  \\\n",
       "movie_id                                                               \n",
       "tt0035423                        0      0                          0   \n",
       "tt0098375                        0      0                          1   \n",
       "tt0098987                        0      0                          0   \n",
       "tt0099005                        0      0                          1   \n",
       "tt0099012                        0      0                          1   \n",
       "\n",
       "           bare-chested-male  friendship  \n",
       "movie_id                                  \n",
       "tt0035423                  0           0  \n",
       "tt0098375                  0           0  \n",
       "tt0098987                  0           0  \n",
       "tt0099005                  0           0  \n",
       "tt0099012                  0           0  \n",
       "\n",
       "[5 rows x 52 columns]"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "movie_df.drop(['story_line', 'keywords'], axis = 1, inplace=True)\n",
    "movie_df.dropna(subset= ['point', 'dollar_budget'], how='any', inplace=True)\n",
    "movie_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "lNMRJgmEaqwT"
   },
   "source": [
    "# Regressors:"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "jOWVhlE-lvnI"
   },
   "source": [
    "## Random Forest:\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "qgjK1c78mLlK"
   },
   "source": [
    "#### RandomizedSearchCV for best hyperparameter values:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "d8tDmjTil2Lg",
    "outputId": "2b47b938-e24f-4a44-db2b-775ff23bb936"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'max_depth': [22, 24, 27, 30, 33],\n",
      " 'max_features': ['auto', 'sqrt'],\n",
      " 'min_samples_leaf': [1, 2, 4, 6],\n",
      " 'min_samples_split': [5, 10, 15],\n",
      " 'n_estimators': [150, 161, 172, 184, 195, 207, 218, 230]}\n"
     ]
    }
   ],
   "source": [
    "n_estimators = [int(x) for x in np.linspace(start = 150, stop = 230, num = 8)]\n",
    "max_depth = [int(x) for x in np.linspace(22, 33, num = 5)]\n",
    "max_features = ['auto', 'sqrt']\n",
    "min_samples_split = [5, 10, 15]\n",
    "min_samples_leaf = [1, 2, 4, 6]\n",
    "\n",
    "random_grid = {'n_estimators': n_estimators,\n",
    "               'max_features': max_features,\n",
    "               'max_depth': max_depth,\n",
    "               'min_samples_split': min_samples_split,\n",
    "               'min_samples_leaf': min_samples_leaf}\n",
    "\n",
    "pprint(random_grid)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "uvzUdMOKmJ49",
    "outputId": "5f09352f-03ce-4ff3-cbb0-6f374dbe42f0"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 10 folds for each of 10 candidates, totalling 100 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  25 tasks      | elapsed:   27.3s\n",
      "[Parallel(n_jobs=-1)]: Done 100 out of 100 | elapsed:  1.2min finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{'n_estimators': 218,\n",
       " 'min_samples_split': 10,\n",
       " 'min_samples_leaf': 1,\n",
       " 'max_features': 'sqrt',\n",
       " 'max_depth': 33}"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(movie_df.drop('point', axis=1), \n",
    "                                                          movie_df['point'], test_size=0.2, random_state=42)\n",
    "rf = RandomForestRegressor()\n",
    "rf_random = RandomizedSearchCV(estimator = rf, param_distributions = random_grid, \n",
    "                               n_iter = 10, cv = 10, verbose=2, random_state=42, \n",
    "                               n_jobs = -1, scoring='neg_mean_squared_error')\n",
    "rf_random.fit(X_train, y_train)\n",
    "rf_random.best_params_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "KKoojcbmmpG7"
   },
   "source": [
    "#### Evaluation method utility:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "id": "fdMvPZCimra6"
   },
   "outputs": [],
   "source": [
    "def pae(y_pred, y_true, a, b):\n",
    "    akbar = len(y_pred)\n",
    "    count = 0\n",
    "    for i in range(len(y_pred)):\n",
    "        target = abs(y_pred[i] - y_true[i])\n",
    "        if (target >= a) and (target < b):\n",
    "            count += 1\n",
    "    return (count/akbar)\n",
    "\n",
    "\n",
    "def evaluate(y_test, y_pred, y_train, y_train_pred):\n",
    "    train_acc = 100 - np.mean(100 * (abs(y_train_pred - y_train) / y_train))\n",
    "    test_acc = 100 - np.mean(100 * (abs(y_pred - y_test) / y_test))\n",
    "    train_r2 = metrics.r2_score(y_train, y_train_pred)\n",
    "    test_r2 = metrics.r2_score(y_test, y_pred)\n",
    "    tr_pae01 = pae(y_train_pred, y_train, 0, 0.5)\n",
    "    tr_pae12 = pae(y_train_pred, y_train, 0, 1)\n",
    "    tr_pae23 = pae(y_train_pred, y_train, 0, 2)\n",
    "    \n",
    "    ts_pa01 = pae(y_pred, y_test, 0, 0.5)\n",
    "    ts_pa12 = pae(y_pred, y_test, 0, 1)\n",
    "    ts_pa23 = pae(y_pred, y_test, 0, 2)\n",
    "    \n",
    "    \n",
    "    train_mae = metrics.mean_absolute_error(y_train, y_train_pred)\n",
    "    test_mae = metrics.mean_absolute_error(y_test, y_pred)\n",
    "    train_rmse = np.sqrt(metrics.mean_squared_error(y_train, y_train_pred))\n",
    "    test_rmse = np.sqrt(metrics.mean_squared_error(y_test, y_pred))\n",
    "    result = pd.DataFrame({\n",
    "    'Data' : ['Train', 'Test'],\n",
    "    'Accuracy' : [train_acc, test_acc],\n",
    "    'R2' : [train_r2, test_r2],\n",
    "    'MAE' : [train_mae, test_mae],\n",
    "    'RMSE'   : [train_rmse, test_rmse],\n",
    "    'PAE(0,0.5)' : [tr_pae01, ts_pa01],\n",
    "    'PAE(0,1)' : [tr_pae12, ts_pa12],\n",
    "    'PAE(0,2)' : [tr_pae23, ts_pa23]\n",
    "    })\n",
    "    return result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "id": "I-13o9LkpvQx"
   },
   "outputs": [],
   "source": [
    "n_est_best = rf_random.best_params_['n_estimators']\n",
    "min_sp_best = rf_random.best_params_['min_samples_split']\n",
    "max_f_best = rf_random.best_params_['max_features']\n",
    "max_d_best = rf_random.best_params_['max_depth']\n",
    "\n",
    "def split_and_fit(input_df):\n",
    "    X_train, X_test, y_train, y_test = train_test_split(input_df.drop('point', axis=1), \n",
    "                                                          input_df['point'], test_size=0.2, random_state=42)\n",
    "    rf = RandomForestRegressor(n_estimators = n_est_best, min_samples_split = min_sp_best,\n",
    "                              max_features = max_f_best, max_depth = max_d_best)\n",
    "    rf.fit(X_train, y_train)\n",
    "    return evaluate(y_test, rf.predict(X_test), y_train, rf.predict(X_train))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "WifT97U3m1oY"
   },
   "source": [
    "#### Handling ouliers:"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "sNCeWG-hm4hd"
   },
   "source": [
    "Let's try something fun. Here we will check three main methods of handling outliers and their impact on model performance:\n",
    "* Using Upper and Lower limit for point. Remove all movies that their point is higher or lower than a specific range.\n",
    "\n",
    "* Using only a Lower limit. Remove all movies that have low budget. We will use budget_cluster and remove the first two groups from."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 110
    },
    "id": "ZRsfUW5Am7ch",
    "outputId": "874af07e-c60f-4715-8515-068434b89ff2"
   },
   "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>Data</th>\n",
       "      <th>Accuracy</th>\n",
       "      <th>R2</th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>PAE(0,0.5)</th>\n",
       "      <th>PAE(0,1)</th>\n",
       "      <th>PAE(0,2)</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Train</td>\n",
       "      <td>93.785538</td>\n",
       "      <td>0.654484</td>\n",
       "      <td>0.372636</td>\n",
       "      <td>0.463168</td>\n",
       "      <td>0.718933</td>\n",
       "      <td>0.970981</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Test</td>\n",
       "      <td>90.682261</td>\n",
       "      <td>0.258011</td>\n",
       "      <td>0.561806</td>\n",
       "      <td>0.686736</td>\n",
       "      <td>0.519177</td>\n",
       "      <td>0.849392</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    Data   Accuracy        R2       MAE      RMSE  PAE(0,0.5)  PAE(0,1)  \\\n",
       "0  Train  93.785538  0.654484  0.372636  0.463168    0.718933  0.970981   \n",
       "1   Test  90.682261  0.258011  0.561806  0.686736    0.519177  0.849392   \n",
       "\n",
       "   PAE(0,2)  \n",
       "0       1.0  \n",
       "1       1.0  "
      ]
     },
     "execution_count": 65,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Second solution\n",
    "upper_lim = movie_df['point'].quantile(.96)\n",
    "lower_lim = movie_df['point'].quantile(.04)\n",
    "no_outliers = movie_df[(movie_df['point'] < upper_lim) & (movie_df['point'] > lower_lim)]\n",
    "split_and_fit(no_outliers)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0, 0.5, 'Frequency')"
      ]
     },
     "execution_count": 66,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(8,5))\n",
    "plt.hist(no_outliers['point'], bins=25, density=True, alpha=0.6, color='g')\n",
    "plt.xlabel(\"Rating\")\n",
    "plt.ylabel(\"Frequency\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "zJMlbfLInDtw"
   },
   "source": [
    "#### Removing unimportant columns based on correlation:\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "kfjdXz5BnE5h"
   },
   "source": [
    "We will drop those columns that their correlation with point is less than 1%, this action will result in a smaller dataset and also a mild increase in accuracy:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "id": "aV6c0YHInG3Q"
   },
   "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>Data</th>\n",
       "      <th>Accuracy</th>\n",
       "      <th>R2</th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>PAE(0,0.5)</th>\n",
       "      <th>PAE(0,1)</th>\n",
       "      <th>PAE(0,2)</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Train</td>\n",
       "      <td>93.579547</td>\n",
       "      <td>0.630955</td>\n",
       "      <td>0.385008</td>\n",
       "      <td>0.47868</td>\n",
       "      <td>0.704657</td>\n",
       "      <td>0.962790</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Test</td>\n",
       "      <td>90.748635</td>\n",
       "      <td>0.270076</td>\n",
       "      <td>0.557276</td>\n",
       "      <td>0.68113</td>\n",
       "      <td>0.516370</td>\n",
       "      <td>0.856876</td>\n",
       "      <td>0.999065</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    Data   Accuracy        R2       MAE     RMSE  PAE(0,0.5)  PAE(0,1)  \\\n",
       "0  Train  93.579547  0.630955  0.385008  0.47868    0.704657  0.962790   \n",
       "1   Test  90.748635  0.270076  0.557276  0.68113    0.516370  0.856876   \n",
       "\n",
       "   PAE(0,2)  \n",
       "0  1.000000  \n",
       "1  0.999065  "
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_corr = no_outliers.corr()\n",
    "a_corrlist = df_corr['point'].sort_values(ascending=False)\n",
    "strong_columns = a_corrlist[abs(a_corrlist) >= 0.01].index.to_list()\n",
    "no_outliers = no_outliers[strong_columns]\n",
    "X_train, X_test, y_train, y_test = train_test_split(no_outliers.drop('point', axis=1), \n",
    "                                                      no_outliers['point'], test_size=0.2, random_state=42)\n",
    "rf = RandomForestRegressor(n_estimators = n_est_best, min_samples_split = min_sp_best,\n",
    "                          max_features = max_f_best, max_depth = max_d_best)\n",
    "rf.fit(X_train, y_train)\n",
    "evaluate(y_test, rf.predict(X_test), y_train, rf.predict(X_train))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Hae2xnm84YnQ"
   },
   "source": [
    "## Lasso and Ridge Regressor:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "id": "8WqhQ5QU4bDW"
   },
   "outputs": [],
   "source": [
    "from sklearn import linear_model\n",
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.preprocessing import PolynomialFeatures"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "id": "yuSTa3bQ5p-a"
   },
   "outputs": [],
   "source": [
    "lasso_params = {'alpha':[0.001, 0.002, 0.005, 0.007 ,0.01, 0.015, 0.02]}\n",
    "ridge_params = {'alpha':[int(x) for x in np.linspace(10, 100, num = 30)]}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "id": "1UI04i1Q62z2"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 10 folds for each of 7 candidates, totalling 70 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed:    0.3s\n",
      "[Parallel(n_jobs=-1)]: Done  70 out of  70 | elapsed:    0.5s finished\n",
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 8 concurrent workers.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 10 folds for each of 10 candidates, totalling 100 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed:    0.2s\n",
      "[Parallel(n_jobs=-1)]: Done 100 out of 100 | elapsed:    0.6s finished\n"
     ]
    }
   ],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(no_outliers.drop('point', axis=1), \n",
    "                                                          no_outliers['point'], test_size=0.2, random_state=42)\n",
    "lasso = linear_model.Lasso()\n",
    "lasso_random = RandomizedSearchCV(estimator = lasso, param_distributions = lasso_params, \n",
    "                               n_iter = 7, cv = 10, verbose=2, random_state=42, \n",
    "                               n_jobs = -1, scoring='neg_mean_squared_error')\n",
    "lasso_random.fit(X_train, y_train)\n",
    "lasso_model = lasso_random.best_estimator_\n",
    "\n",
    "\n",
    "ridge = linear_model.Ridge()\n",
    "ridge_random = RandomizedSearchCV(estimator = ridge, param_distributions = ridge_params, \n",
    "                               n_iter = 10, cv = 10, verbose=2, random_state=42, \n",
    "                               n_jobs = -1, scoring='neg_mean_squared_error')\n",
    "ridge_random.fit(X_train, y_train)\n",
    "ridge_model = ridge_random.best_estimator_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "LC1M8VvD9qg2"
   },
   "source": [
    "#### Lasso and Ridge Performance without pipeline:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "id": "w2xhNmvI8tMy"
   },
   "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>Data</th>\n",
       "      <th>Accuracy</th>\n",
       "      <th>R2</th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>PAE(0,0.5)</th>\n",
       "      <th>PAE(0,1)</th>\n",
       "      <th>PAE(0,2)</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Train</td>\n",
       "      <td>90.702471</td>\n",
       "      <td>0.222575</td>\n",
       "      <td>0.557772</td>\n",
       "      <td>0.694759</td>\n",
       "      <td>0.519541</td>\n",
       "      <td>0.842031</td>\n",
       "      <td>0.998128</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Test</td>\n",
       "      <td>90.434103</td>\n",
       "      <td>0.207389</td>\n",
       "      <td>0.576768</td>\n",
       "      <td>0.709776</td>\n",
       "      <td>0.504210</td>\n",
       "      <td>0.821328</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    Data   Accuracy        R2       MAE      RMSE  PAE(0,0.5)  PAE(0,1)  \\\n",
       "0  Train  90.702471  0.222575  0.557772  0.694759    0.519541  0.842031   \n",
       "1   Test  90.434103  0.207389  0.576768  0.709776    0.504210  0.821328   \n",
       "\n",
       "   PAE(0,2)  \n",
       "0  0.998128  \n",
       "1  1.000000  "
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Lasso:\n",
    "evaluate(y_test, lasso_model.predict(X_test), y_train, lasso_model.predict(X_train))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "id": "YdgdQ0vT9x_G"
   },
   "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>Data</th>\n",
       "      <th>Accuracy</th>\n",
       "      <th>R2</th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>PAE(0,0.5)</th>\n",
       "      <th>PAE(0,1)</th>\n",
       "      <th>PAE(0,2)</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Train</td>\n",
       "      <td>90.696707</td>\n",
       "      <td>0.222360</td>\n",
       "      <td>0.558095</td>\n",
       "      <td>0.694855</td>\n",
       "      <td>0.520477</td>\n",
       "      <td>0.840861</td>\n",
       "      <td>0.998362</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Test</td>\n",
       "      <td>90.424114</td>\n",
       "      <td>0.206217</td>\n",
       "      <td>0.577415</td>\n",
       "      <td>0.710301</td>\n",
       "      <td>0.507016</td>\n",
       "      <td>0.821328</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    Data   Accuracy        R2       MAE      RMSE  PAE(0,0.5)  PAE(0,1)  \\\n",
       "0  Train  90.696707  0.222360  0.558095  0.694855    0.520477  0.840861   \n",
       "1   Test  90.424114  0.206217  0.577415  0.710301    0.507016  0.821328   \n",
       "\n",
       "   PAE(0,2)  \n",
       "0  0.998362  \n",
       "1  1.000000  "
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Ridge:\n",
    "evaluate(y_test, ridge_model.predict(X_test), y_train, ridge_model.predict(X_train))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "ujiuctzj95nA"
   },
   "source": [
    "#### Adding pipeline:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "id": "4EEK5kK49OBQ"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\hoomo\\miniconda3\\lib\\site-packages\\sklearn\\model_selection\\_search.py:282: UserWarning: The total space of parameters 7 is smaller than n_iter=10. Running 7 iterations. For exhaustive searches, use GridSearchCV.\n",
      "  % (grid_size, self.n_iter, grid_size), UserWarning)\n",
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 8 concurrent workers.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 10 folds for each of 7 candidates, totalling 70 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done  25 tasks      | elapsed:   41.8s\n",
      "[Parallel(n_jobs=-1)]: Done  70 out of  70 | elapsed:  1.7min finished\n",
      "C:\\Users\\hoomo\\miniconda3\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:531: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 966.2747727631747, tolerance: 0.26530341259068574\n",
      "  positive)\n",
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 8 concurrent workers.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 10 folds for each of 10 candidates, totalling 100 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done  25 tasks      | elapsed:    1.9s\n",
      "[Parallel(n_jobs=-1)]: Done 100 out of 100 | elapsed:    7.0s finished\n"
     ]
    }
   ],
   "source": [
    "lasso_params = {'fit__alpha':[0.001, 0.002, 0.005, 0.007 ,0.01, 0.015, 0.02]}\n",
    "ridge_params = {'fit__alpha':[int(x) for x in np.linspace(10, 100, num = 30)]}\n",
    "\n",
    "pipe_lasso = Pipeline([('poly', PolynomialFeatures()),\n",
    "                 ('fit', linear_model.Lasso())])\n",
    "lasso_random = RandomizedSearchCV(estimator = pipe_lasso, param_distributions = lasso_params, \n",
    "                               n_iter = 10, cv = 10, verbose=2, random_state=42, \n",
    "                               n_jobs = -1, scoring='neg_mean_squared_error')\n",
    "lasso_random.fit(X_train, y_train)\n",
    "lasso_model = lasso_random.best_estimator_\n",
    "\n",
    "pipe_ridge = Pipeline([('poly', PolynomialFeatures()),\n",
    "                 ('fit', linear_model.Ridge())])\n",
    "ridge_random = RandomizedSearchCV(estimator = pipe_ridge, param_distributions = ridge_params, \n",
    "                               n_iter = 10, cv = 10, verbose=2, random_state=42, \n",
    "                               n_jobs = -1, scoring='neg_mean_squared_error')\n",
    "ridge_random.fit(X_train, y_train)\n",
    "ridge_model = ridge_random.best_estimator_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "_MwPYab0_NCT"
   },
   "source": [
    "#### Lasso and Ridge Performance with pipeline:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "id": "AXmVwEd7_Or1"
   },
   "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>Data</th>\n",
       "      <th>Accuracy</th>\n",
       "      <th>R2</th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>PAE(0,0.5)</th>\n",
       "      <th>PAE(0,1)</th>\n",
       "      <th>PAE(0,2)</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Train</td>\n",
       "      <td>91.088085</td>\n",
       "      <td>0.289001</td>\n",
       "      <td>0.534574</td>\n",
       "      <td>0.664415</td>\n",
       "      <td>0.536859</td>\n",
       "      <td>0.858413</td>\n",
       "      <td>0.99883</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Test</td>\n",
       "      <td>90.614276</td>\n",
       "      <td>0.244337</td>\n",
       "      <td>0.564435</td>\n",
       "      <td>0.693035</td>\n",
       "      <td>0.518241</td>\n",
       "      <td>0.837231</td>\n",
       "      <td>1.00000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    Data   Accuracy        R2       MAE      RMSE  PAE(0,0.5)  PAE(0,1)  \\\n",
       "0  Train  91.088085  0.289001  0.534574  0.664415    0.536859  0.858413   \n",
       "1   Test  90.614276  0.244337  0.564435  0.693035    0.518241  0.837231   \n",
       "\n",
       "   PAE(0,2)  \n",
       "0   0.99883  \n",
       "1   1.00000  "
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Lasso:\n",
    "evaluate(y_test, lasso_model.predict(X_test), y_train, lasso_model.predict(X_train))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "id": "3X5e9s6E98uk"
   },
   "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>Data</th>\n",
       "      <th>Accuracy</th>\n",
       "      <th>R2</th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>PAE(0,0.5)</th>\n",
       "      <th>PAE(0,1)</th>\n",
       "      <th>PAE(0,2)</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Train</td>\n",
       "      <td>91.261008</td>\n",
       "      <td>0.314430</td>\n",
       "      <td>0.524043</td>\n",
       "      <td>0.652425</td>\n",
       "      <td>0.547859</td>\n",
       "      <td>0.865668</td>\n",
       "      <td>0.998830</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Test</td>\n",
       "      <td>90.590588</td>\n",
       "      <td>0.239701</td>\n",
       "      <td>0.565833</td>\n",
       "      <td>0.695158</td>\n",
       "      <td>0.526660</td>\n",
       "      <td>0.839102</td>\n",
       "      <td>0.999065</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    Data   Accuracy        R2       MAE      RMSE  PAE(0,0.5)  PAE(0,1)  \\\n",
       "0  Train  91.261008  0.314430  0.524043  0.652425    0.547859  0.865668   \n",
       "1   Test  90.590588  0.239701  0.565833  0.695158    0.526660  0.839102   \n",
       "\n",
       "   PAE(0,2)  \n",
       "0  0.998830  \n",
       "1  0.999065  "
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Ridge:\n",
    "evaluate(y_test, ridge_model.predict(X_test), y_train, ridge_model.predict(X_train))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "k1aoaFCiOnE0"
   },
   "source": [
    "## SVRs:"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "CLwDv599OqEB"
   },
   "source": [
    "### SVR - linear:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "id": "7Pn41tdIOttB"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 10 folds for each of 10 candidates, totalling 100 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  25 tasks      | elapsed:   20.5s\n",
      "[Parallel(n_jobs=-1)]: Done 100 out of 100 | elapsed:  1.2min finished\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>Data</th>\n",
       "      <th>Accuracy</th>\n",
       "      <th>R2</th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>PAE(0,0.5)</th>\n",
       "      <th>PAE(0,1)</th>\n",
       "      <th>PAE(0,2)</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Train</td>\n",
       "      <td>87.059856</td>\n",
       "      <td>-0.592052</td>\n",
       "      <td>0.809045</td>\n",
       "      <td>0.994222</td>\n",
       "      <td>0.367891</td>\n",
       "      <td>0.660192</td>\n",
       "      <td>0.963492</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Test</td>\n",
       "      <td>86.669700</td>\n",
       "      <td>-0.651187</td>\n",
       "      <td>0.839423</td>\n",
       "      <td>1.024447</td>\n",
       "      <td>0.351731</td>\n",
       "      <td>0.620206</td>\n",
       "      <td>0.958840</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    Data   Accuracy        R2       MAE      RMSE  PAE(0,0.5)  PAE(0,1)  \\\n",
       "0  Train  87.059856 -0.592052  0.809045  0.994222    0.367891  0.660192   \n",
       "1   Test  86.669700 -0.651187  0.839423  1.024447    0.351731  0.620206   \n",
       "\n",
       "   PAE(0,2)  \n",
       "0  0.963492  \n",
       "1  0.958840  "
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.svm import SVR\n",
    "\n",
    "svr_l_params = {'kernel': ['linear'],\n",
    "                'gamma': [10, 1, 0.1, 0.01, 0.001],\n",
    "                'C': [0.1, 1, 10, 50]}\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(no_outliers.drop('point', axis=1), \n",
    "                                                          no_outliers['point'], test_size=0.2, random_state=42)\n",
    "svr_linear_random = RandomizedSearchCV(estimator = SVR(), param_distributions = svr_l_params, \n",
    "                               n_iter = 10, cv = 10, verbose=2, random_state=42, \n",
    "                               n_jobs = -1, scoring='neg_mean_squared_error')\n",
    "svr_linear_random.fit(X_train, y_train)\n",
    "svr_linear_model = svr_linear_random.best_estimator_\n",
    "evaluate(y_test, svr_linear_model.predict(X_test), y_train, svr_linear_model.predict(X_train))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "hKVgjwklXexf"
   },
   "source": [
    "### SVR - RBF:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "id": "FtKWkYRaV3v6"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 10 folds for each of 10 candidates, totalling 100 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  25 tasks      | elapsed:   15.6s\n",
      "[Parallel(n_jobs=-1)]: Done 100 out of 100 | elapsed:  1.0min finished\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>Data</th>\n",
       "      <th>Accuracy</th>\n",
       "      <th>R2</th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>PAE(0,0.5)</th>\n",
       "      <th>PAE(0,1)</th>\n",
       "      <th>PAE(0,2)</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Train</td>\n",
       "      <td>91.498215</td>\n",
       "      <td>0.312924</td>\n",
       "      <td>0.505145</td>\n",
       "      <td>0.653141</td>\n",
       "      <td>0.582729</td>\n",
       "      <td>0.867072</td>\n",
       "      <td>0.998362</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Test</td>\n",
       "      <td>90.511667</td>\n",
       "      <td>0.224661</td>\n",
       "      <td>0.567190</td>\n",
       "      <td>0.702000</td>\n",
       "      <td>0.521048</td>\n",
       "      <td>0.838167</td>\n",
       "      <td>0.999065</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    Data   Accuracy        R2       MAE      RMSE  PAE(0,0.5)  PAE(0,1)  \\\n",
       "0  Train  91.498215  0.312924  0.505145  0.653141    0.582729  0.867072   \n",
       "1   Test  90.511667  0.224661  0.567190  0.702000    0.521048  0.838167   \n",
       "\n",
       "   PAE(0,2)  \n",
       "0  0.998362  \n",
       "1  0.999065  "
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "svr_r_params = {'kernel': ['rbf'],\n",
    "                'gamma': [10, 1, 0.1, 0.01, 0.001],\n",
    "                'C': [0.1, 1, 10, 50]}\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(no_outliers.drop('point', axis=1), \n",
    "                                                          no_outliers['point'], test_size=0.2, random_state=42)\n",
    "svr_rbf_random = RandomizedSearchCV(estimator = SVR(), param_distributions = svr_r_params, \n",
    "                               n_iter = 10, cv = 10, verbose=2, random_state=42, \n",
    "                               n_jobs = -1, scoring='neg_mean_squared_error')\n",
    "svr_rbf_random.fit(X_train, y_train)\n",
    "svr_rbf_model = svr_rbf_random.best_estimator_\n",
    "evaluate(y_test, svr_rbf_model.predict(X_test), y_train, svr_rbf_model.predict(X_train))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'kernel': 'rbf', 'gamma': 0.01, 'C': 1}"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "svr_rbf_random.best_params_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "OA9izaTDYiKi"
   },
   "source": [
    "### SVR - Poly:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "id": "n5ELGmsoYly3"
   },
   "outputs": [],
   "source": [
    "# svr_p_params = {'kernel': ['poly'],\n",
    "#                 'gamma': [10, 1, 0.1, 0.01, 0.001],\n",
    "#                 'C': [0.1, 1, 10]}\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(no_outliers.drop('point', axis=1), \n",
    "                                                          no_outliers['point'], test_size=0.2, random_state=42)\n",
    "# svr_poly_random = RandomizedSearchCV(estimator = SVR(), param_distributions = svr_p_params, \n",
    "#                                n_iter = 10, cv = 10, verbose=2, random_state=42, \n",
    "#                                n_jobs = -1, scoring='neg_mean_squared_error')\n",
    "\n",
    "\n",
    "# svr_poly_random.fit(X_train, y_train)\n",
    "# svr_poly_model = svr_poly_random.best_estimator_\n",
    "\n",
    "# svr_poly_model = SVR(kernel = 'poly', gamma= 0.01, C= 1)\n",
    "# svr_poly_model.fit(X_train, y_train)\n",
    "# evaluate(y_test, svr_poly_model.predict(X_test), y_train, svr_poly_model.predict(X_train))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "UkEGD5PNRV5g"
   },
   "source": [
    "## KNN regressor:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "id": "93ZVHrqzRYoB"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 10 folds for each of 10 candidates, totalling 100 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed:    0.8s\n",
      "[Parallel(n_jobs=-1)]: Done 100 out of 100 | elapsed:    2.2s finished\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>Data</th>\n",
       "      <th>Accuracy</th>\n",
       "      <th>R2</th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>PAE(0,0.5)</th>\n",
       "      <th>PAE(0,1)</th>\n",
       "      <th>PAE(0,2)</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Train</td>\n",
       "      <td>100.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Test</td>\n",
       "      <td>90.072851</td>\n",
       "      <td>0.136399</td>\n",
       "      <td>0.600753</td>\n",
       "      <td>0.74088</td>\n",
       "      <td>0.482694</td>\n",
       "      <td>0.792329</td>\n",
       "      <td>0.998129</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    Data    Accuracy        R2       MAE     RMSE  PAE(0,0.5)  PAE(0,1)  \\\n",
       "0  Train  100.000000  1.000000  0.000000  0.00000    1.000000  1.000000   \n",
       "1   Test   90.072851  0.136399  0.600753  0.74088    0.482694  0.792329   \n",
       "\n",
       "   PAE(0,2)  \n",
       "0  1.000000  \n",
       "1  0.998129  "
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.neighbors import KNeighborsRegressor\n",
    "knn_params = {  'n_neighbors': [int(x) for x in np.linspace(2, 50, num = 20)],\n",
    "                'weights': ['uniform', 'distance'],\n",
    "                'algorithm' : ['auto', 'ball_tree', 'kd_tree', 'brute']}\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(no_outliers.drop('point', axis=1), \n",
    "                                                          no_outliers['point'], test_size=0.2, random_state=42)\n",
    "knn_random = RandomizedSearchCV(estimator = KNeighborsRegressor(), param_distributions = knn_params, \n",
    "                               n_iter = 10, cv = 10, verbose=2, random_state=42, \n",
    "                               n_jobs = -1, scoring='neg_mean_squared_error')\n",
    "knn_random.fit(X_train, y_train)\n",
    "knn_model = knn_random.best_estimator_\n",
    "evaluate(y_test, knn_model.predict(X_test), y_train, knn_model.predict(X_train))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "McJakNW0cGHU"
   },
   "source": [
    "## Util:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "id": "4SSjToV3cCZ9"
   },
   "outputs": [],
   "source": [
    "upper_lim = movie_df['point'].quantile(.96)\n",
    "lower_lim = movie_df['point'].quantile(.04)\n",
    "no_outliers = movie_df[(movie_df['point'] < upper_lim) & (movie_df['point'] > lower_lim)]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "zSdLmrfhbPd-"
   },
   "source": [
    "## XGboost:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {
    "id": "B3ax5hnX0kOW"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 10 folds for each of 25 candidates, totalling 250 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  25 tasks      | elapsed:   11.3s\n",
      "[Parallel(n_jobs=-1)]: Done 146 tasks      | elapsed:  1.6min\n",
      "[Parallel(n_jobs=-1)]: Done 250 out of 250 | elapsed:  2.9min finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[09:19:19] WARNING: src/objective/regression_obj.cu:152: reg:linear is now deprecated in favor of reg:squarederror.\n"
     ]
    }
   ],
   "source": [
    "import xgboost as xgb\n",
    "\n",
    "data_dmatrix = xgb.DMatrix(data=no_outliers.drop('point', axis=1),label=no_outliers['point'])\n",
    "X_train, X_test, y_train, y_test = train_test_split(no_outliers.drop('point', axis=1), \n",
    "                                                          no_outliers['point'], test_size=0.2, random_state=42)\n",
    "\n",
    "xgb_params = {  'n_estimators': [int(x) for x in np.linspace(10, 1000, num = 40)],\n",
    "                'max_depth': [int(x) for x in np.linspace(4, 40, num = 10)],\n",
    "                'learning_rate ' : [0.0001, 0.001, 0.01, 0.1, 1.0],\n",
    "                'colsample_bytree' : [0.4, 0.3, 0.2, 0.1],\n",
    "                'subsample' : [0.5, 0.3, 1]}\n",
    "\n",
    "\n",
    "xgb_random = RandomizedSearchCV(estimator = xgb.XGBRegressor(), param_distributions = xgb_params, \n",
    "                               n_iter = 25, cv = 10, verbose=2, random_state=42, \n",
    "                               n_jobs = -1, scoring='neg_mean_squared_error')\n",
    "\n",
    "xgb_random.fit(X_train, y_train)\n",
    "xgb_model = xgb_random.best_estimator_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "id": "ywpbIxRP-p_2"
   },
   "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>Data</th>\n",
       "      <th>Accuracy</th>\n",
       "      <th>R2</th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>PAE(0,0.5)</th>\n",
       "      <th>PAE(0,1)</th>\n",
       "      <th>PAE(0,2)</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Train</td>\n",
       "      <td>92.120776</td>\n",
       "      <td>0.438043</td>\n",
       "      <td>0.471809</td>\n",
       "      <td>0.590686</td>\n",
       "      <td>0.612450</td>\n",
       "      <td>0.902410</td>\n",
       "      <td>0.999766</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Test</td>\n",
       "      <td>90.712709</td>\n",
       "      <td>0.261838</td>\n",
       "      <td>0.559423</td>\n",
       "      <td>0.684963</td>\n",
       "      <td>0.513564</td>\n",
       "      <td>0.849392</td>\n",
       "      <td>0.998129</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    Data   Accuracy        R2       MAE      RMSE  PAE(0,0.5)  PAE(0,1)  \\\n",
       "0  Train  92.120776  0.438043  0.471809  0.590686    0.612450  0.902410   \n",
       "1   Test  90.712709  0.261838  0.559423  0.684963    0.513564  0.849392   \n",
       "\n",
       "   PAE(0,2)  \n",
       "0  0.999766  \n",
       "1  0.998129  "
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "evaluate(y_test, xgb_model.predict(X_test), y_train, xgb_model.predict(X_train))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "id": "wYtLTB8dAAya"
   },
   "outputs": [],
   "source": [
    "# xgb_random.best_params_\n",
    "new_params = {'colsample_bytree': 0.4, 'learning_rate ': 1.0,\n",
    "              'max_depth': 4, 'subsample': 1}\n",
    "cv_results = xgb.cv(dtrain=data_dmatrix, params=new_params, nfold=10,\n",
    "      num_boost_round=50,early_stopping_rounds=10,metrics=[\"rmse\", \"mae\"], as_pandas=True, seed=123)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "id": "__9YciDbAtjD"
   },
   "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>train-mae-mean</th>\n",
       "      <th>train-mae-std</th>\n",
       "      <th>train-rmse-mean</th>\n",
       "      <th>train-rmse-std</th>\n",
       "      <th>test-mae-mean</th>\n",
       "      <th>test-mae-std</th>\n",
       "      <th>test-rmse-mean</th>\n",
       "      <th>test-rmse-std</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>0.509905</td>\n",
       "      <td>0.002154</td>\n",
       "      <td>0.633271</td>\n",
       "      <td>0.002255</td>\n",
       "      <td>0.549105</td>\n",
       "      <td>0.017383</td>\n",
       "      <td>0.681246</td>\n",
       "      <td>0.016753</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    train-mae-mean  train-mae-std  train-rmse-mean  train-rmse-std  \\\n",
       "24        0.509905       0.002154         0.633271        0.002255   \n",
       "\n",
       "    test-mae-mean  test-mae-std  test-rmse-mean  test-rmse-std  \n",
       "24       0.549105      0.017383        0.681246       0.016753  "
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cv_results.tail(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "id": "vBVw86R8B3T0"
   },
   "outputs": [],
   "source": [
    "# print(\"Train Accuracy: \", (0.571710 / cv_results.tail(1)['train-mae-mean'].values[0]) * 90.466982)\n",
    "# print(\"Test Accuracy: \", (0.572226 / cv_results.tail(1)['test-mae-mean'].values[0]) * 90.503220)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "do5jfXy1hLvx"
   },
   "source": [
    "## Gradient Boost:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "id": "-mQLyE5nhQHr"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 10 folds for each of 30 candidates, totalling 300 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  25 tasks      | elapsed:    4.1s\n",
      "[Parallel(n_jobs=-1)]: Done 146 tasks      | elapsed:   25.0s\n",
      "[Parallel(n_jobs=-1)]: Done 300 out of 300 | elapsed:   48.9s finished\n"
     ]
    }
   ],
   "source": [
    "from sklearn.ensemble import GradientBoostingRegressor\n",
    "X_train, X_test, y_train, y_test = train_test_split(no_outliers.drop('point', axis=1), \n",
    "                                                          no_outliers['point'], test_size=0.2, random_state=42)\n",
    "\n",
    "gb_params = {  'n_estimators': [int(x) for x in np.linspace(10, 100, num = 20)],\n",
    "                'max_depth': [int(x) for x in np.linspace(3, 25, num = 10)],\n",
    "                'learning_rate' : [0.5, 0.04, 0.05, 0.01],\n",
    "                'min_samples_split' : [1, 2, 5, 10],\n",
    "                'min_samples_leaf' : [5, 8, 10],\n",
    "                'max_features' : [int(x) for x in np.linspace(1, X_train.shape[1], num = 20)]}\n",
    "\n",
    "gb_random = RandomizedSearchCV(estimator = GradientBoostingRegressor(), param_distributions = gb_params, \n",
    "                               n_iter = 30, cv = 10, verbose=2, random_state=42, \n",
    "                               n_jobs = -1, scoring='neg_mean_squared_error')\n",
    "\n",
    "gb_random.fit(X_train, y_train)\n",
    "gb_model = gb_random.best_estimator_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {
    "id": "QWcgYYu7mZIz"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'n_estimators': 81,\n",
       " 'min_samples_split': 5,\n",
       " 'min_samples_leaf': 10,\n",
       " 'max_features': 22,\n",
       " 'max_depth': 7,\n",
       " 'learning_rate': 0.04}"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gb_random.best_params_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "id": "JkgVSU3ZjJly"
   },
   "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>Data</th>\n",
       "      <th>Accuracy</th>\n",
       "      <th>R2</th>\n",
       "      <th>MAE</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>PAE(0,0.5)</th>\n",
       "      <th>PAE(0,1)</th>\n",
       "      <th>PAE(0,2)</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Train</td>\n",
       "      <td>92.411490</td>\n",
       "      <td>0.480301</td>\n",
       "      <td>0.453084</td>\n",
       "      <td>0.568043</td>\n",
       "      <td>0.631875</td>\n",
       "      <td>0.916452</td>\n",
       "      <td>0.999766</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Test</td>\n",
       "      <td>90.777012</td>\n",
       "      <td>0.270418</td>\n",
       "      <td>0.555067</td>\n",
       "      <td>0.680971</td>\n",
       "      <td>0.517306</td>\n",
       "      <td>0.858746</td>\n",
       "      <td>0.999065</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    Data   Accuracy        R2       MAE      RMSE  PAE(0,0.5)  PAE(0,1)  \\\n",
       "0  Train  92.411490  0.480301  0.453084  0.568043    0.631875  0.916452   \n",
       "1   Test  90.777012  0.270418  0.555067  0.680971    0.517306  0.858746   \n",
       "\n",
       "   PAE(0,2)  \n",
       "0  0.999766  \n",
       "1  0.999065  "
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "evaluate(y_test, gb_model.predict(X_test), y_train, gb_model.predict(X_train))"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [
    "gZahqOxcmL_E",
    "N7NYSLSyqX-M"
   ],
   "name": "Rating Analysis.ipynb",
   "provenance": []
  },
  "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
}