FinalProject_group050.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# COGS 108 - Final Project\n",
    "# TEAM GROUP: 50\n",
    "# TEAM NAME: Movie Stars"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Names and PID\n",
    "- Anthony Ostia &emsp; XXXXXXXXX\n",
    "- Ramin Atrian &emsp; XXXXXXXXX\n",
    "- Chuyang Xiao &emsp; XXXXXXXXX\n",
    "- Yifeng Zhang &emsp; XXXXXXXXX\n",
    "- Ian Nelson &emsp; XXXXXXXXX"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Introduction\n",
    "In this final project of Data science, we are trying to generate a recommendation system of movies for IMDB users with the respect of their historical ratings. At first, after obtaining the datasets about both film's final ratings and individual users' ratings, we would try to combine them together with movieID as the key for further comparisons. Next, by visualizing the collected data, we would study the correlation between individual as well as final ratings of films and film genres. At last, we would use the model we generate to recommend films for each individual user. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Background \n",
    "Since the beginning of the 20th century, film has been an important recreation to human beings possessing both artistic and commercial significance. Moreover, in the Information Age, rather than being overwhelmed by the gigantic informative flow, film won unprecedented attention from the population with the aid of widespread accessibility to the Internet. Just like Hsu (2014) argued in his essay, in the era of Web 2.0, film critics are no longer privileges limited to professionals yet typical rights to common audiences as well, due to the introduction of film rating websites such as IMDB or Rotten Tomatoes: each audience could comment or grade films online and then share their evaluations and opinions to the public. Consequently, the data collected thereby is not only abundant in quantity, but also diverse in nature, including millions of users’ personal information as well as preference of films. Hence, considering it a perfect dataset to study users’ attributes to films, we would like to focus on the ratings on movie websites in this project.\n",
    "\n",
    "As increasing audiences begin to consider online ratings as critical criteria to decide whether or  not to see a film, a number of data scientists struggle to explore the rating data, which has provided us with the direction of further research. In the research accomplished by Hsu (2014) focusing on predicting the movies user rating based on the IMDB attributes, by analyzing the ratings and critics of 31506 movies, they generated a mathematical model mimicking the manners of the preference of general IMDB users, and hence predicted the final rating of a newly coming movie via entering related parameters to it. Regardingly, we found that instead of being purely independent, the IMDB ratings of new films could be roughly predicted by past data. Furthermore, in Breuss’s research (2012), the personal information on users’ social media, including gender, hobby, or regular activities, was taken into account and therefore, predicted the IMDB movie ratings they might give, which inspires that individuals’ statistics could also be of concern. \n",
    "\n",
    "All in all, in this data science project, we will combine these two studies and step forward: we will study what kind of movies will a user like based on the historical ratings and review texts of that user on movies he/she has seen previously. From our perspective, this study is not only interesting but also quite applicable, could possibly promote the recommendation system in movie rating websites.\n",
    "\n",
    "Reference:\n",
    "1. Hsu, P. Y., Shen, Y. H., & Xie, X. A. (2014, October). Predicting movies user ratings with imdb attributes. In International Conference on Rough Sets and Knowledge Technology (pp. 444-453). Springer, Cham.\n",
    "2. Oghina, A., Breuss, M., Tsagkias, M., & De Rijke, M. (2012, April). Predicting imdb movie ratings using social media. In European Conference on Information Retrieval (pp. 503-507). Springer, Berlin, Heidelberg."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Research Question\n",
    "How to recommend films in their favorite genres to each individual IMDB user according to their historical ratings?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Hypothesis\n",
    "Within this data science group project, we would like to develope a recommendation systems for the movie rating websites which involves to predict the users' preference in movie according to their rating history and then provide them with certain genre of movies that they favour. Hence, we generated a couple of hypothesises with the respect of this topic:\n",
    "1. With the respect of the users, one would like to rate higher on the genres (higher than the average) he or she loves, which indicates his or her love. On the opposite, to the genres he or she dislikes, his or hers ratings would be much lower(lower than or equal to the average).\n",
    "2. Moreover, if a user loves a certain genre of film, he or she tends to watch more this genre of films.\n",
    "3. If a user loves a certain genre of film, this genre of film must exist among the films he or she rates the highest"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data Description\n",
    "\n",
    "In this group project, we would majorly use two datasets, the rating.csv and movie.csv\n",
    "\n",
    "Dataset name: rating.csv\n",
    "\n",
    "Reference: https://www.kaggle.com/rounakbanik/the-movies-dataset/version/7#\n",
    "\n",
    "Number of observations: 80001052\n",
    "\n",
    "We would only use the rating.csv in this data package. It includes 138493 IMDB users' ratings on films in the past 15 years, and each observation refers to one rating. This dataset would construct the rating history of these users that we are going to study and generate predictive model. The columns of this dataset are:\n",
    "\n",
    "1. userID: the IMDB ID of this user that made this rating\n",
    "2. movieID: the movieID of the film got rated in this rating\n",
    "3. rating: the rating the user gives to the film\n",
    "4. timestamp: the time this rating is made\n",
    "\n",
    "In further explorations, we would possibly focus on the information in userID, movieID and rating column.\n",
    "\n",
    "Dataset name: movie.csv\n",
    "\n",
    "Reference: https://www.kaggle.com/grouplens/movielens-20m-dataset#tag.csv\n",
    "\n",
    "Number of observations: 27277\n",
    "\n",
    "In this dataset, each observation refers to the information of one film in the IMDB collection. The columns of this dataset are:\n",
    "1. movieID: the movieID of the film, and the reference to the user's ratings\n",
    "2. title: the title of the film\n",
    "3. genres: the genres of the film\n",
    "\n",
    "In further explorations, we would utilize the information in all three columns.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<br>\n",
    "<br>\n",
    "<br>\n",
    "<br>\n",
    "<br>\n",
    "<br>\n",
    "<br>\n",
    "<br>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Setup"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "#import the useful modules and packages.\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import seaborn as sns\n",
    "import math\n",
    "import matplotlib.pyplot as plt\n",
    "from collections import defaultdict\n",
    "from collections import Counter\n",
    "from sklearn.metrics import precision_recall_fscore_support"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# load in our datasets in csv with pandas method.\n",
    "df_rating = pd.read_csv('rating.csv') \n",
    "df_movie = pd.read_csv('movie.csv',dtype = {'title':str, 'genres':str, 'movieId':int})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>userId</th>\n",
       "      <th>movieId</th>\n",
       "      <th>rating</th>\n",
       "      <th>timestamp</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>3.5</td>\n",
       "      <td>2005-04-02 23:53:47</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>29</td>\n",
       "      <td>3.5</td>\n",
       "      <td>2005-04-02 23:31:16</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>32</td>\n",
       "      <td>3.5</td>\n",
       "      <td>2005-04-02 23:33:39</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>47</td>\n",
       "      <td>3.5</td>\n",
       "      <td>2005-04-02 23:32:07</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>50</td>\n",
       "      <td>3.5</td>\n",
       "      <td>2005-04-02 23:29:40</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   userId  movieId  rating            timestamp\n",
       "0       1        2     3.5  2005-04-02 23:53:47\n",
       "1       1       29     3.5  2005-04-02 23:31:16\n",
       "2       1       32     3.5  2005-04-02 23:33:39\n",
       "3       1       47     3.5  2005-04-02 23:32:07\n",
       "4       1       50     3.5  2005-04-02 23:29:40"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_rating.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>movieId</th>\n",
       "      <th>title</th>\n",
       "      <th>genres</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>Toy Story (1995)</td>\n",
       "      <td>Adventure|Animation|Children|Comedy|Fantasy</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>Jumanji (1995)</td>\n",
       "      <td>Adventure|Children|Fantasy</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>Grumpier Old Men (1995)</td>\n",
       "      <td>Comedy|Romance</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>Waiting to Exhale (1995)</td>\n",
       "      <td>Comedy|Drama|Romance</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>Father of the Bride Part II (1995)</td>\n",
       "      <td>Comedy</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   movieId                               title  \\\n",
       "0        1                    Toy Story (1995)   \n",
       "1        2                      Jumanji (1995)   \n",
       "2        3             Grumpier Old Men (1995)   \n",
       "3        4            Waiting to Exhale (1995)   \n",
       "4        5  Father of the Bride Part II (1995)   \n",
       "\n",
       "                                        genres  \n",
       "0  Adventure|Animation|Children|Comedy|Fantasy  \n",
       "1                   Adventure|Children|Fantasy  \n",
       "2                               Comedy|Romance  \n",
       "3                         Comedy|Drama|Romance  \n",
       "4                                       Comedy  "
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_movie.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data Cleaning / Pre-processing"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We drop the timestamp column from the df_rating dataframe. from since it is not beneficial for our project. Then, we drop any data with missing information from the df_rating and df_movie dataframes. Next, we change the scale of the ratings from the rating columns from five to ten by multiplying each rating by 2. Since the information is too large, we decided to create a new dataframe called train that will use the first 2000 users from the df_ratings dataframe. We will then remove any rows that do not have any ratings in the train dataframe. We then create a dataframe called test which takes the last five rows of the train dataframe and drops the index values.Then for both the train and test dataframe, we will create three columns called average score, genre, and variance. In the average score column, we add up all the values for the movie being rated and then divide it by the total amount of ratings for that movie. In the genre column, we put the genre corresponding to the proper movie. Lastly in the variance column, we put the difference between a rating of a movie and the average rating for that movie."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# drop the timestamp column\n",
    "df_rating = df_rating.drop(columns = ['timestamp'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# drop data with missing values\n",
    "df_rating.dropna(inplace=True)\n",
    "df_movie.dropna(inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# convert the original rating in the scale of 5 into the rating in the scale of 10.\n",
    "df_rating['rating'] *= 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# The original ratings dataset is too large\n",
    "# we would select the first 2000 users as our training set\n",
    "train = df_rating[df_rating['userId']<=2000]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# remove rows without ratings\n",
    "rows_to_drop = []\n",
    "for n in range(len(train)):\n",
    "    if train['movieId'][n] not in df_movie['movieId']:\n",
    "        rows_to_drop.append(n)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# split the train and test dataset\n",
    "train = train.drop(rows_to_drop)\n",
    "frames = [train[train['userId'] == i+1][-5:] for i in range(2000)]\n",
    "test = pd.concat(frames)\n",
    "todrop = test.index.values\n",
    "train = train.drop(todrop)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# calculate the average ratings\n",
    "average_ratings = {}\n",
    "for index, row in train.iterrows():\n",
    "    if row[1] in average_ratings:\n",
    "        average_ratings[row[1]].append(row[2])\n",
    "    else:\n",
    "        average_ratings[row[1]] = [row[2]]\n",
    "        \n",
    "average_dict = {}\n",
    "for key,value in average_ratings.items():\n",
    "    average_dict[key] = sum(value)/len(value)\n",
    "    \n",
    "train['average_score'] = [average_dict[row[1]] for index,row in train.iterrows()]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# append genres to train dataset\n",
    "genre = []\n",
    "for i in train['movieId']:\n",
    "    g = \"\"\n",
    "    for val in df_movie[df_movie['movieId'] == i]['genres']:\n",
    "        g = val\n",
    "    genre.append(g)\n",
    "train['genres'] = genre"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# calculate variance for each user on each movie\n",
    "train['variance'] = train.apply(lambda row: row.rating - row.average_score, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "# calculate average score for the test set\n",
    "average_ratings = {}\n",
    "for index, row in test.iterrows():\n",
    "    if row[1] in average_ratings:\n",
    "        average_ratings[row[1]].append(row[2])\n",
    "    else:\n",
    "        average_ratings[row[1]] = [row[2]]\n",
    "\n",
    "average_dict = {}\n",
    "for key,value in average_ratings.items():\n",
    "    average_dict[key] = sum(value)/len(value)\n",
    "\n",
    "test['average_score'] = [average_dict[row[1]] for index,row in test.iterrows()]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# append genres to the test set\n",
    "genre = []\n",
    "for i in test['movieId']:\n",
    "    g = \"\"\n",
    "    for val in df_movie[df_movie['movieId'] == i]['genres']:\n",
    "        g = val\n",
    "    genre.append(g)\n",
    "test['genres'] = genre"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "test['variance'] = test.apply(lambda row: row.rating - row.average_score, axis=1)\n",
    "test['recommend'] = test.apply(lambda row: row.rating >= 6.0, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>userId</th>\n",
       "      <th>movieId</th>\n",
       "      <th>rating</th>\n",
       "      <th>average_score</th>\n",
       "      <th>genres</th>\n",
       "      <th>variance</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>7.0</td>\n",
       "      <td>6.484076</td>\n",
       "      <td>Adventure|Children|Fantasy</td>\n",
       "      <td>0.515924</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>29</td>\n",
       "      <td>7.0</td>\n",
       "      <td>7.825397</td>\n",
       "      <td>Adventure|Drama|Fantasy|Mystery|Sci-Fi</td>\n",
       "      <td>-0.825397</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>32</td>\n",
       "      <td>7.0</td>\n",
       "      <td>7.763955</td>\n",
       "      <td>Mystery|Sci-Fi|Thriller</td>\n",
       "      <td>-0.763955</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>47</td>\n",
       "      <td>7.0</td>\n",
       "      <td>8.200632</td>\n",
       "      <td>Mystery|Thriller</td>\n",
       "      <td>-1.200632</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>50</td>\n",
       "      <td>7.0</td>\n",
       "      <td>8.773723</td>\n",
       "      <td>Crime|Mystery|Thriller</td>\n",
       "      <td>-1.773723</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   userId  movieId  rating  average_score  \\\n",
       "0       1        2     7.0       6.484076   \n",
       "1       1       29     7.0       7.825397   \n",
       "2       1       32     7.0       7.763955   \n",
       "3       1       47     7.0       8.200632   \n",
       "4       1       50     7.0       8.773723   \n",
       "\n",
       "                                   genres  variance  \n",
       "0              Adventure|Children|Fantasy  0.515924  \n",
       "1  Adventure|Drama|Fantasy|Mystery|Sci-Fi -0.825397  \n",
       "2                 Mystery|Sci-Fi|Thriller -0.763955  \n",
       "3                        Mystery|Thriller -1.200632  \n",
       "4                  Crime|Mystery|Thriller -1.773723  "
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>userId</th>\n",
       "      <th>movieId</th>\n",
       "      <th>rating</th>\n",
       "      <th>average_score</th>\n",
       "      <th>genres</th>\n",
       "      <th>variance</th>\n",
       "      <th>recommend</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>169</th>\n",
       "      <td>1</td>\n",
       "      <td>8482</td>\n",
       "      <td>7.0</td>\n",
       "      <td>8.000000</td>\n",
       "      <td>Drama|Fantasy|Horror</td>\n",
       "      <td>-1.000000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>170</th>\n",
       "      <td>1</td>\n",
       "      <td>8507</td>\n",
       "      <td>10.0</td>\n",
       "      <td>8.500000</td>\n",
       "      <td>Crime|Drama|Horror</td>\n",
       "      <td>1.500000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>171</th>\n",
       "      <td>1</td>\n",
       "      <td>8636</td>\n",
       "      <td>9.0</td>\n",
       "      <td>7.242105</td>\n",
       "      <td>Action|Adventure|Sci-Fi|IMAX</td>\n",
       "      <td>1.757895</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>172</th>\n",
       "      <td>1</td>\n",
       "      <td>8690</td>\n",
       "      <td>7.0</td>\n",
       "      <td>7.000000</td>\n",
       "      <td>Comedy|Drama|Sci-Fi|War</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>173</th>\n",
       "      <td>1</td>\n",
       "      <td>8961</td>\n",
       "      <td>8.0</td>\n",
       "      <td>7.568889</td>\n",
       "      <td>Action|Adventure|Animation|Children|Comedy</td>\n",
       "      <td>0.431111</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     userId  movieId  rating  average_score  \\\n",
       "169       1     8482     7.0       8.000000   \n",
       "170       1     8507    10.0       8.500000   \n",
       "171       1     8636     9.0       7.242105   \n",
       "172       1     8690     7.0       7.000000   \n",
       "173       1     8961     8.0       7.568889   \n",
       "\n",
       "                                         genres  variance  recommend  \n",
       "169                        Drama|Fantasy|Horror -1.000000       True  \n",
       "170                          Crime|Drama|Horror  1.500000       True  \n",
       "171                Action|Adventure|Sci-Fi|IMAX  1.757895       True  \n",
       "172                     Comedy|Drama|Sci-Fi|War  0.000000       True  \n",
       "173  Action|Adventure|Animation|Children|Comedy  0.431111       True  "
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "# save the train and test dataframe for future use\n",
    "train.to_csv('train.csv')\n",
    "test.to_csv('test.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data Visualization"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##   Graph 1 Scatter matrix of the dataframe"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 9 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = pd.plotting.scatter_matrix(train[['rating','average_score','variance']])\n",
    "plt.suptitle('scatter-matrix of the dataframe')\n",
    "fig = plt.gcf()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "-The upper left cell of this scatter-matrix is a histogram of the rating given in each user review. Our 1-to-10 scoring system is a result of multiplying ratings on a five-star scale by two. A notable consequence of the original rating system which can be seen in this histogram is that odd ratings in our 1-to-10 system (which correspond to ratings with half-stars in the original system) are less common than even ratings (whole stare ratings). The histogram also shows that the distribution of ratings is skewed to the left. While the odd ratings are less common than the even ones, their shape follows the general shape of the even numbered ratings.\n",
    "-The middle left cell shows that films of a broad range of average scores are given a wide range of ratings, though high ratings are rarely given to films with very low average scores.\n",
    "-The center cell is a histogram of the number of ratings films with different ranges of average score have received. This indicates that films with higher average scores tend to be rated by more people.\n",
    "-The lower left cell shows a plot of film rating vs difference between that rating and the film’s average rating. It shows that the range of variance for lower ratings is greater than the range of variance for higher ratings. This makes sense considering the skewed distribution of ratings observed in the upper left cell.\n",
    "-The bottom middle cell shows that variance is simply the difference between average score and rating: it is simply a transformation of the top middle graph, which is just a transposition of the middle left graph.\n",
    "-The bottom right graph is a histogram of variance values. It is skewed to the left, which is to be expected given that the distribution of ratings is skewed to the left."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Graph 2 Average rating of each genre"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "score_dict = {}\n",
    "for index, row in train.iterrows():\n",
    "    genres = row[4].split('|')\n",
    "    for gen in genres:\n",
    "        if gen in score_dict:\n",
    "            score_dict[gen].append(row[3])\n",
    "        else:\n",
    "            score_dict[gen] = [row[3]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX4AAAFQCAYAAAC4fBW6AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjAsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+17YcXAAAgAElEQVR4nOydfZxVVbn4v8+8MCMM6gwQSvJW3kqHFxWsxEzNyFKjN6trb+ZVu2KQ1fVmvtyrvdGve6VbcsuprKwMLLGSzHxBxC6QFQM1Y74UpgOFEgijMAQzwPP7Y+097HM4Z2avvdfM2efM+n4++zOz1zn7Oc/aL89e61nPepaoKh6Px+MZOlSVWgGPx+PxDC7e8Hs8Hs8Qwxt+j8fjGWJ4w+/xeDxDDG/4PR6PZ4jhDb/H4/EMMbzh9wxpROQaEbmlBL87VkR+JSI7RWThIP3mGSLy18H4LU+2qSm1Ap7BQ0RWAtOBo1R1b4nVGXRE5AzgNlU9JixT1QUlUucjwDbgcPWTaTyDjG/xDxFEZBJwGqDAnAH6jZI1JMRQTvfzROAxb/RLe98MVcrpQfGk40PAI8CtwIVhoYi8RkSeE5HqSNk7RKQt+L9KRD4tIk+JyPMi8mMRaQo+myQiKiIXi8hGYEVQfkcg84XAndEckT1KRH4uIi+KyO9E5PMisiry+atE5AER2S4iT4rIe4pVSERWisgXRGQ1sBt4mYhcJCKPBy6Uv4jIvwbfHQH8EhgnIruCbZyI3CAit+XV50IR2Sgi20Tk2sjvHSYi3xORHcFvfKov14mIzArq+ELwd1ZQHl6DTwV6vLHAsXUicmOgxxYRaRGRw4LPGkXkbhHZGuhyt4gcEzm2SUS+KyKbg89/lif730Tk7yLyrIhc1If+kyPuqOUi8rXwXAWfv1ZE1ohIp4j8IehRRa/N50RkdXD8/SIyOu885983ReV5HKOqfhsCG7ABuByYAfQAYyOfPQXMjuzfAXw6+P8KzAvjGKAO+AawJPhsEqYH8X1gBHBYUP4vwMjg+18Bfh+RfXuwDQeOBzYBq4LPRgT7F2HckCdi3CHHF6nTSmAj0Bx8vxY4F3g5IMDpmBfCScH3zwD+mifjBoz7J1qfbwGHYdxie4Hjgs//H/Aw0Bicj7Z8eRG5TcAO4IOBbhcE+6OCz28FPt/H9fofYFkgZyTwc+CLwWejgHcF53BkcL1+Fjn2F8CPAj1rgdMj9d8HfDYoPyc4P41FdPg1cCMwDHgd8GLkXL0UeD6QUQXMDvbHRK7NU8ArgnO5Evh/xe6b/uT5zbE9KLUCfhuEi2we2h5gdLD/BPCJyOefB74T/D8S6AImBvuPA2dFvnt0IKsm8gC/rI/fPjL4zhFAdXDsK/N+OzT87wX+L+/4bwDXF5G9EvhsP3X/GXBF8P8ZxDP8x0Q+/y3wz8H/fwHOjnx2Sb68yGcfBH6bV/Zr4MPB/7dSxPBjXlpdwMsjZacATxf5/gnAjsj1OUABYx7U/x9ATaTs78BrC3x3AuYlMTxSdlvkXF0F/CDvmPuACyPX5rrIZ5cD9+ad55dFPu9Tnt/cbt63NjS4ELhfVbcF+4uDsv+J7K8RkbnAO4F1qtoRfDYR+KmIHIjI2w+MjexvCv8JXEZfAN4NjMEYIYDRmJZdTfT7ef9PBF4jIp2RshrgB33ULXo8IvIW4HpMS7MK0ypu7+P4QjwX+X830BD8P47iuuczDujIK+vAtGz7YwxG71YRCcsE8+JERIZjrt2bMa16gJHBuR8PbFfVHUVkP6+q+yL70frl679dVXdHyjYF8sFcq3eLyFsjn9cCD0X2i53HqLyQOPI8jvCGv8IJ/MLvAapFJHwQ64AjRWS6qv5BVR8TkQ7gLcD7MC+CkE3Av6jq6gKyJwX/Rgco3we8DXgj8Aympb8DY7i2YlqRxwB/Cr4/PnLsJuBhVZ1tUcXe3xaROuBOzHjGXaraE/i3Jf+7CXkWo/tjwf74Pr67GWPMokwA7o3xO9swLfNmVf1bgc//DXgl8BpVfU5ETgDWY+q5CWgSkSNVtbPAsXF5NpAzPGL886/VD1T10hS/Eb0eLuR5YuIHdyuft2Na6MdjXAInAMcB/4cxkCGLMf7812N8xiEtwBdEZCKAiIwRkbf18XsjMX7x5zGt1t5wSVXdD/wEuEFEhovIq/J0uBt4hYh8UERqg+1kETkuZl2HYV5qW4F9Qev/TZHPtwCjROSImPLy+TFwdTC4+lJgXh/fvQdTl/eJSI2IvBdzDe7u70dU9QBmnOF/ROQlACLyUhE5O/jKSMyLoVPMQPv1kWOfxQxifz3Qs1ZEXm9b0aDHtxZzrYaJyClAtDV+G/BWETlbRKpFpF7MPIFjCgrsH9fyPH3gDX/lcyHwXVXdqKrPhRvwv8D75WAo3RLMYOiKiEsI4KuYQcb7RWQnZqD3NX383vcxLo2/YVrGj+R9Pg/TC3gO48JZgnlRoKo7MYb6nzEt5ueAL2GMeb8Ex38MY6B3YHofyyKfPxH83l+CyJFxceRG+CzwV+BpYDmwNNS9gC7PA+dhWufPA58Czss7t31xFWZA/hEReTH4vVcGn30F4zbbhjm/+b2ID2LGUp7A+PA/HvM383k/ZmzhecxYzI84eK02YXp212BetJuAfyehTXEtz9M3EgyieDwlQUS+hJlQdmG/X84YwZjIP6vq6aXWZTAQkR8BT6jq9f1+2ZNp/NvUM6iIidOfJoZXAxcDPy21XnEQkaNF5FQxcxteiWnNl4XuSQjcbC8P6vtmTIv8Z/0d58k+fnDXM9iMxLhbxmF87guBu0qqUXyGYcJLJwOdmPkIXy+pRgPLUZgxmVEYF9dcVV1fWpU8LvCuHo/H4xlieFePx+PxDDG84fd4PJ4hRln4+EePHq2TJk0qtRoej8dTVrS2tm5T1TH55WVh+CdNmsTatWtLrYbH4/GUFcGM/EPwrh6Px+MZYnjD7/F4PEMMb/g9Ho9niOENv8fj8QwxBszwi8h3guXdHo2UNYlZVu/Pwd/GvmTE/J1DtlIxf/586uvrERHq6+uZP3++tYws1Qfc1Km2tjanPrW1tQOgaXyqq6tz9Kmuru7/oDxcXScXcirxnhk1alROfUaNGjUAmg4uS5YsYcqUKVRXVzNlyhSWLFmSSI6TczNQK7xg0vueBDwaKfsvDi7p92ngS3FkzZgxQwuByeet1dXVunLlSq2uru4tG2zmzZunNTU1unDhQu3q6tKFCxdqTU2Nzps3L7aMUHdAp0+fnrNfClzUqaamRgFtbGzUtrY2bWxsVEBramoGUPPiVFVVKaANDQ3a2tqqDQ0NCmhVVVVsGdHrctNNNyW+TtHjbr/99kRyoscsXbq0Iu6ZpqYmBbS5uVk7Ojq0ublZAW1qahpAzQeWxYsX6+TJk3XFihXa3d2tK1as0MmTJ+vixYut5NieG2CtFrLPhQpdbZgl1qKG/0ng6OD/o4En48jpy/BXV1fnlIXGf7Cpq6vThQsX5pQtXLhQ6+rqYsso9MCW8iF2VafGxsacstD4l4LQ6EcJjb+NDBfXyYWcSr1nmpubc8pCA1euNDc364oVK3LKVqxYcUg9+8P23GTF8HdG/pfofoFjP4JZCGLthAkTip6ElStX5pStXLmyJDcIoF1dXTllXV1d1g/x9OnTc8rCln8SffK3JDJc1KmtrS2nrK2traR1am1tzSlrbW21rtNNN92UUxa2/G11uf3223PKwpa/jYylS5fmlIUtf1uydM90dHTklHV0dJS14a+qqtLu7u6csu7ubquepqr9ucmc4Q/2d8SR41v8yeuT5tistvjTHutb/P3LTIpv8RdmqLf4nbt6QuPvffzFZSYlqz7+NHXyPv54MpPiffyFGeo+/v8md3D3v+LIKWb4g4ql7p66Yt68eVpXV6eA1tXVWd3sIa7rk/Z4F3UKjX+4pR3YTVun0PiHm213O9TBxXVyIacS75nQwIVbORv9kMWLF2tzc7NWVVVpc3OztdEPsTk3xQz/gOXjF5ElwBnAaMyCG9djVu/5MTABsy7re1R1e3+yZs6cqT5XTzJEhIG6xqWiEuuUJfz5rRxEpFVVZ+aXD1iSNlW9oMhHZw3Ub3o8Ho+nf/zMXY/H4xlieMPv8Xg8Qwxv+D2eEuIifYRnaOEiJYY3/B5PiaiurubAgQM0NDTQ2tpKQ0MDBw4c8MbfU5T58+fT0tLCggUL6OrqYsGCBbS0tFgb/wGL6nGJj+pJTiVGaFRKnUSEhoYGdu7c2Vs2cuRIdu3aVdL6Vcr5rUTq6+tZsGABn/zkJ3vLvvzlL3PNNdewZ8+eQ75fLKqnrFv8hbISljozYVoqsU6ViKtrNGLEiBwZI0aMcKKLv2cqk71793LZZZfllF122WXs3bvXSk5ZG/7ohITofjlTqD7lXqdKxNV9t2XLFmbNmsXmzZuZNWsWW7ZscaKLv2cqk7q6OlpaWnLKWlpaqKurs5JTFoutezyVTFtbG88++yxtbW2lVsWTcS699FKuuuoqwLT0W1pauOqqqw7pBfRHxfj4K80v6ao+lXZeIFt1SqNL6JKJHh/uJ5Hp75lsM23aNNrb23v3p06dmuhlP3/+fL71rW+xd+9e6urquPTSS1m0aFHB71akj99THmTJB50lXQBOOeWUHPfMKaecUjJdPANHvtEHaG9vZ9q0adayFi1axJ49e1BV9uzZU9To94U3/J4BJ0s+6CyNC40fP541a9Zw6qmn8uyzz3LqqaeyZs0axo8fXxJ9PANHaPSj4znR8sHG+/g9nhKxceNGJkyYwJo1axg3bhxgXgYbN24ssWaegWDmzJmsXr0agNWrV3PyySdTqjB1b/g9nhLijfzQobGxsc/9wcS7ejwej1Nqa2tzxk9qa2tLrVImeOCBB3jb297Gtm3beNvb3sYDDzxQMl284XdIlgYNPZ5SUFtby759+2hsbKStrY3Gxkb27ds35I1/OG6zbNkyxowZw7Jly3LKBxvv6nFIOEjow+E8Q5XQ6G/fbtZX2r59O01NTezYsaPEmpWWcDxn06ZNvWWlHM/xLX58S70v/LkZWCrx/D788MN97scha2G3Lti4cWNONFlSo+/ivHjDj7vp95WIPzcDSyWe39NPP73P/ThkKQQ4a7i4Z7zh93g8zqipqWHHjh00NTXR3t7e6+apqfFe5Szhr4bH43FGT08PtbW17Nixo3dWak1NDT09PSXWzBPFG36Px+MUb+Szj3f1eDwezxDDG36Px+MZYnjD7/F4PEMM7+P3eDxOKRRXXqpQzCzpkiW84fd4PE7J0gz2LOmSJbyrx+PxeIYY3vB7PB7PEMMbfo/H4xlieMPv8Xg8Qwxv+D0ej2cAaWpqOiSTZn52zaampkHVyUf1eDwezwCyY8eOfiOKBjvltG/xezwezxDDt/g9Ho9nANHrD4cbjuj/O4OIN/weT5lSaEnDfJdBdBlET2mQz7wYy9WjNwyOPlAiwy8inwAuARRoBy5S1T2l0MXjKVey6Dv2lAeD7uMXkZcCHwNmquoUoBr458HWo1LJYgSBx+NJR/5zXejZtnmuS+XqqQEOE5EeYDiwuUR6VByV2Arsz6Xh3RmeSsf1cz3ohl9V/yYiNwIbgX8A96vq/fnfE5GPAB8BmDBhwuAq6ckU/d305fYic0UWBw0rkWL3VzknfZPBVl5EGoE7gfcCncAdwFJVva3YMTNnztS1a9f2Jzf1hXCVwa+UusQ5LonsLNdpIM/VYMhIKifL19qVnCzpklSOi+uUVIaItKrqzPzvliKO/43A06q6VVV7gJ8As0qgR6bwvnmPxzNYlMLHvxF4rYgMx7h6zgL6bs4PASrRN+/xeLJJKXz8vxGRpcA6YB+wHvjmYOvh8Xg8g0V/jbbGxsZB0sRQkpQNqnq9qr5KVaeo6gdVdW8p9PB4PNkiSy5PVyGUqpqzFSob7Kg0P3PXM2D4maUeW7Lk8sySLq4pyyRtriczVIouWSN8cPra8l8MHo9n4CnLFn+W3sRZ0sVTHhTqCYGflOYpjus5G2Vp+D0eW7JkbH1jwWOL60RvZenq8XhsceV2ytLgo6c4/V2noX6NfIvfcwhZah1nDd9aLw98mo++8YbfcwhZM279+Td9Ppp0uEiC5xsL5YU3/BnBJ9wqTn/+zcFexKLScNE6zlpjwdM3Q9rwZynO3NXgjX+BeDye/hjShr8SWylZXObN4ylHKrkRNaQNv8fj8RSjkhtR/Rp+EXk58FdV3SsiZwDTgO+raudAK+fxePoma8m/0lLJrey0uLzWcVr8dwIzReRYTBbNu4DFwDmxf8Xj8TinwKIbZb0qFAzeWFe5vTwKnZM01zuO4T+gqvtE5B3AIlVdJCLrE/1aBeJbKB5P9p6DLEaC1dfXs3fv3t7fr6urY8+ePYOrREAcw98jIhcAFwJvDcpqB06l8qKS/YCegSFrRtIF/jnom6jRD9m7dy/19fUlMf5xDP9FwGXAF1T1aRGZDPxgYNXqm0p8cCoRf50K443k0CPf6PdXPtAM+mLrSchfbN3ZItP9GKWD33sh1e+U23eypEuc72RJlzjfydoi6ZV2fuN8ZzB1Cb8Hub76QmU2xNSv4GLrcaJ62oF86S9g1sn9vKo+b6FrpvAtL48tvhfjScrw4cMP2d+9e3dJdInj6vklsB8TyQPwz8Bw4DngVg76/T2eisc3FjxJ2b17N0cddRQrV67kjDPOKJnRh3iG/42qelJkv11E1qnqSSLygYFSzFM6KrFVW4l18pQfW7Zs4bjjjiu1GrEMf7WIvFpVfwsgIicD1cFn+wZMsyFIVibjVGKrthLr5CkfVLXg812qMdY4hv8S4Dsi0gAI8CJwiYiMAL44kMoNJSpxMo7HU0r6akjFbURlpTHmmn4Nv6r+DpgqIkcE+9EQlx8PlGKe5FTizeriIfYMHVw0pFzOli12/5aqgRcnqqcOeBcwCaiJhCB9dkA18ySiEnsOlVgnz9Bk1qxZLF26lPPPP581a9aUTI84rp67MOGbrUBpZht4PJ4Bw1Vem0rsabpmzZo1jBs3rtRqxDL8x6jqmwdcE49nCJElI+kir81guEVCyvkFMmfOHL797W9z8cUXs2zZspLpEcfwrxGRqaraPuDaeDxDANeZFiuJSnfrLVu2jCuvvLKkRh/iGf7XAR8Wkacxrh4BVFWnDahmHo/HU4F873vfK7UKsQz/WwZcizKnkrumHo+n8qgq9oGIhCM6O4tsHkzXNH/LLx+Mxdqzioj0ufmXomcoMWfOHLZu3cqcOXNKqkdfLf7FwHmYaB7FuHhCFHjZAOrVL76VnX0q3V/r8dhw7LHHsmzZMsaMGdO7v2HDhpLoUtTwq+p5wd/Jg6dOPPzgmMfjKTeee+65HBs1cuTIkulS1NUTIiIPxinzeDweT2GqqqrYtWsXI0eOZN26dYwcOZJdu3ZRVdWvCR4Qirb4RaQek355tIg0ctDVczjw0kHQzePxeCqC/fv3U11dza5du5gxYwZgXgb79+8viT59+fj/Ffg4MA7j5w8N/4vA/w6wXh6Pc/y4kKeUlMrIF6IvH/9Xga+KyHxVXeTyR0XkSOAWYApmoPhfVPXXLn/Dk45KM5J+XMhTasoqLbOqLhKRKcDxQH2k/PspfverwL2qer6IDMO4lEpCpRk4F3gj2Tf+nvHYEt4ztbW1PPTQQ5x55pn09PRkOjvn9cAZGMN/D2ZC1yogkeEP0ju/HvgwgKp2A91JZKXFhxt6bPH3jCcptbW1dHcbU9fd3c2wYcPo6emxlhNteCRdsD3OkPL5wFnAc6p6ETAd6HsNu76ZDGwFvisi60XklmBRlxxE5CMislZE1m7dujXFz3k8Hk9yohMOo/u2PPTQQ33ux9XFprwYcQz/P1T1ALAvmM37d2C81a/kUgOcBNysqicCXcCn87+kqt9U1ZmqOjOc8ODxeDyDTaHZ+Ul6eWeeeWaf+4NJHMO/NhiM/RYmumcdkGYg9q/AX1X1N8H+UsyLwOPxJMBFa9Qz8PT09DBs2DBWr16d2M0TUltby6pVq6itrU10fJ8+fjF30BdVtRNoEZF7gcNVtS3RrwGq+pyIbBKRV6rqkxg30mNJ5Xk8Q5msLennKUy42HpPTw+ve93rcsqTEB0rSPKi77PFr0areyL7z6Qx+hHmAz8UkTbgBGCBA5kej8eTWVy4i0JEhF/84heJe3dx0jKvE5GTg0XXnaCqvwdmupLn8Qx1okbEu3qGBuedd17iY+P4+F8D/FpEnhKRNhFpD1rqHo8nI+RHnqQ5vpJSZ1fq+MeA+vgDzk4k2ePxlAWVOjehEsc/XI0VxJm522Gvnsfj8XgGAhcvrTgtfo/H4/FkBBc5f0qTDNrj8Tjn1ltvLbUKgPE/R/3qSf3QnkOJGv3p06cXLI9DrBa/iEwE/klVl4vIYUCNqvp1dz2ejCAiHHXUUSX3X9fW1rJv376csn379lFbW5tqwpInl7RRXHFW4LoUM7v2G0HRMcDPrH/J4/EMGKrKm9/85pIPWuYb/f7KPclIG8UVx9XzUeBUzAIsqOqfgZck+jWPxzMkmDt3Lp2dncydO7fUqlQkzc3NdHR00NzcnOh46a+FICK/UdXXiMh6VT1RRGqAdao6LdEvJmDmzJm6du3aPr/joovrqpucRo7rxRqyUKcs6uJKTql16avFl+YeTHv/FnJFJK2fy2cppNQ9o6TY1klEWlX1kMmycVr8D4vINcBhIjIbuAP4eXxVPXFxlXLV4yk1aV0RnoEljuH/NCZ/fjtmHd57gOsGUimPxxOfY489Nif/y7HHHltqlTwDSENDQ871bmhosJYRZwLXAUxK5m8l0NHj8QwwGzZs8C3rIcSuXbuYMmUK99xzD+eccw67du2ylhFn6cV2zILoUV4A1gKfV9XnrX/V4/FULCeddBLr16/vTS9w4oknsm7dulKrVTHU1tbyxz/+kYkTJ/bu24bKxnH1/BL4BfD+YPs5xug/B9xq9WsejyeHSkwitm7dut6BRlX1Rj8gnNgWbkkmtk2dOvUQI9/T08PUqVOt5MQx/G9U1atVtT3YrgVOV9UvAZOsfs3j8fTicjB/7NixPP7444wdOzatWs5YsMAvsxESTmxrbGykra2NxsbG3oltNrS3t1uVFyPOzN1qEXm1qv4WQEROBqqDz/ysDI8nA2zZsoXjjjuu1GrkcM0115RaBSAbaxWERn/79u0AbN++naamJnbs2JFI3oDP3AUuAb4tIk+LyDPAt4FLRWQE8EXrX/R4PAPChRdeWGoVMklWQksffvjhPvfjsnTp0j7349DvBK7eL4ocAaCqL1j/SkqGygSurE3GcS0nS7q4klPqCUZZu2dc65MlXdIQLmgTtviB3ha/jS62E+TSTOBCRM7FxPBfISL/KSL/GVtTj8cz4GTRx+/nExykpqaGHTt20NTURHt7e6/Rr6lJlhlfRLjzzjsHLlePiLQA78UskC7Au4GJiX7N40lBJUbAuCL08W/ZsqXUqvSyYcOGxMdOmzYt51pPmzZoGWIGhJ6enl7jP23atF6jbxuGGW3Vn3/++QXL4xCnxT9LVT8E7FDVzwCnAK+w+hWPJyWhEQijIMK/3vhnkzRrwk6bNo329nbmzJkDwJw5c2hvb09k/GtqanJmuSZtYbugp6cnR5ekaaqjMsLNljiGf0/wd7eIjAN6gKOtf2kI4FukA0/4sPjc7tkmXBM2yXUKQxOXLVuW89c2ZBFMNM3kyZN56qmnmDx5sk8PHRDH8P9cRI4E/htYBzwDLB5IpcqRrCVYq6+vz3kJ1dfXl0QPz+CQpvWXVaIx72l45plnOPbYY3nmmWfcKFYB9Gn4RaQKeFBVO1X1Toxv/1Wq6gd3++C0004r6e/X19ezd+/e3oG+sWPHsnfvXm/8K5ishCy6orq6mu3btzN16lS2b99OdXV1/wflUewlWEkvx6T0afiDBG1fi+zvLUU4ZzmhqvzqV78q6c21d+9egN6BvvBvWD7YeBeYx5b9+/fnvMz279+fSE5+KgPb1AZZZP78+b09+vr6eubPn28tI46r50EReZf4pzUWWWp5zZkzh61bt/YOkpWCrLnAPEOH6CBx+BwkHSTOCvPnz6elpYUFCxbQ1dXFggULaGlpsTf+hUaI8/yFO4EDmEHdF4P9F/s7zuU2Y8YM7Q9TlXSkkYHJYFpwG0wZoZza2tqcstra2tT1S3qcqzq5kJMvMy2Vcs/ky0xzLKBVVVW6fPlyraqqSlynxsZGbWtr08bGxsT3zJw5c3LK5syZ4+S6l4q6ujpduHBhTtnChQu1rq6u4PeBtVrApsaeuVtK+pq563KpwlLPEnQ10zCUM3bsWFauXMkZZ5zR6+4Z7HPjuk5p5eTLTHv/Z+WeiX4/zVKH4fGlrlNVVRUHDhzoLQv3be+ZrVu3Mnr06N6ybdu2MWbMmLL184sIXV1dDB8+vLds9+7djBgxwu3MXTF8QET+I9gfLyKvTqe+G7wboTgikjOpx5+TykZEuOKKKyrmOjc1NeWMCzU1NSWSM2bMmBz365gxY1yqOejU1dXR0tKSU9bS0kJdXZ2VnDg+/q9jJm29L9jfRWTA15M9pk6diqrm+DZVtSIGtjzFuemmm0qtQi+TJk1iw4YNTJo0yfrYESNGsG3bNiZOnMiGDRuYOHEi27ZtY8SIEYn1ueWWWxIfmyUuvfRSrrrqKr785S+ze/duvvzlL3PVVVdx6aWX2gkq5P+JbsC64O/6SNkf+jvO5VbMx49j/2bS41zp4rI+dXV1OccX8wHa6Jb0OBd1cn2tQ5lpqaR7JiozzbGHH354jh7hflxEJMevT+DvFxFrXVyfmywwb9683ue7rq5O582bV/S7FPHxx2nx94hIdXDCEJExmMFeT0aZNm3aIaGbe/fuLetoBk9htIivulj5YLBz584+9/tDVXn66adzDFW4b8vSpUtz5CRJYZw1Fi1axJ49e1BV9uzZw6JFi6xlxDH8NwE/BV4iIl8AVgEVtbROpcWZh1PbZ82axebNm5k1a1ZOuaeyKNSiK7U+fe33h4hw9dVX55RdffXViZ7L888/P8fHH01sNpTp1/Cr6g+BT2EWXXkWeLuq3jHQig0WAw/jmL8AACAASURBVDFA/LWvpR8CSbt60cyZM1m9ejVHH300q1evZubMQwb2PZ5MMnv2bG6++WYuv/xyXnjhBS6//HJuvvlmZs+eXWrVKodCrYW8lsNNmAydg+bTz98G0sfvQkYWdZk9e3ZO2ezZsxP5NrNwXlzKKSYrKWmPdVUnl2ShTm9605tURBRQEdE3velNJdOlnCGFj78VuE5EnhKRG0XESdNRRKpFZL2I3O1CnieXBx54IKeL+8ADD5RapUxQiSHAVVVVOde6qirW+kqZ5v777+91Eakq999/f0n1Ofvss3vPc1VVFWeffba1jCVLljBlyhSqq6uZMmUKS5YsSaRL9Fondk0XehsU2oAm4FLgQeDPcY/rQ94nMVk+7+7vu4PV4j/66KMrosUftpTyt1JERbiqU9bkROUlxYUu4bWur6/XRx55ROvr6xNd63y90hwbbl/60pdS3zO33Xabk2v9ta99LVXvA9C5c+dqZ2enzp07VwGrXsjixYt18uTJumLFCu3u7tYVK1bo5MmTdfHixYnrtHTp0n7rRJEWv42hfjWwENgA/DzucUVkHRO8QN6QJcNfaoMSfr+5uVk7Ojq0ubm57I1klnQZaDm2uDq/9fX1OWWh8U9KqV9mhb5fSsMvIjp37tycsrlz51q9XJubm3XFihU5ZStWrNDm5mYrXWzPTWLDD/wX8GfgXuAi4Mj+jokhcykwAzijmOEHPgKsBdZOmDChz5NQKYbJtS4NDQ3a2tqqDQ0NJa+Ti7wrlXqd0uryyCOP5JQ98sgjZW/4b7vttpyysOU/2LqEcjo7O3PKOjs7reRUVVVpd3d3Tll3d7dWVVVZ67J06dKcsrDlX+T7iQ3/vwKj+/te3A04D/h68H9Rwx/dhlqL35Uu/ZUNhj5ZOr+VWCfwLf6B1EV1iLb4zbE0Ylw9rw+3OMcVkfVF4K+YlbyeA3YDt/V1jDf8yXTJgsvIVZ1qamoKyqipqSnbOrmQk2Ufv4vz68rHn+Y6DUkfP3AJ0A7sAB4C/gGs6O+4OBu+xZ9ZXbJWp2KyXMgo9+uUP5ifxuiHeqU5dqDOTal0UXUTXrp48WJtbm7WqqoqbW5utjb6ITb1KWb448R9XQGcDHSo6pnAiUBnjOM8CUm7gEqxcL5yDvML003knxufhoLedMXhFk1nXApOPPHEHH1OPPFEaxmFjFUSqqqqcg1ewmfgvvvu6z3PBw4c4L777rOWccEFF/Doo4+yf/9+Hn30US644IJEurg4NzUxvrNHVfcE8aJ1qvqEiLzSXt1DUdWVwEoXsiqJZcuWpUofu3//fqqrqw/JZ550+bosEKabyD83Pg1F9ujq6upzfzCZPn16r/EXEU444QTWr19fMn2yQpzX319F5EjgZ8ADInIX0DGwannSsn///pwWQamN/ty5c+ns7GTu3Lkl1SOLiAj33ntv4klk1dXVOZN5kixM7pI//elPTJkyhY0bNzJlyhT+9Kc/lUyX9evXc9lll9HZ2clll13mjX6A1QpcInI6cARwr6p2D5hWeRRbgSuLq15lQRdXZKlOoZxZs2axdOlSzj//fNasWZNYThp9snTPhD27hoYGHn74YU4//XR27dqVqoeXZgWuCRMmsGnTpkPKx48fz8aNGxPJTErWnqdSkHgFriiq+rCqLhtMo+9JxrRp03JagZXiC1+zZg3jxo3rNfpDnQMHDlBdXU1PTw8zZsygp6fnEDffYLJx48ZDDK6IDLrRd838+fOpr69HRKivr7df3Nwho0aNynm2R40aZS2jfEf78rjyyitLrUIvt99+e0l/f9q0abS3t+cMhLa3t6cy/u94xzscaujJJ8lKVSH79+9nwYIFdHV1sWDBgpK69UaNGoWq0tzcTEdHB83NzahqIuPkirFjx/L4448zduzYRMfPnz+flpaWnHPc0tJSEuM/atQotm/fnnN+t2/fbn9+C40QZ23z4Zz2usyZMyenbM6cOWVfp6zIyZoutbW1OWW1tbWpQzLTHJs/KSmcRzLYhOcyP/7eVpe6ujpduHBhTtnChQtTr2qXBNvzS5FwTisff6nwPn57XbZu3cro0aN7y7Zt28aYMWPKuk5ZkZNFXfJ9/La65MtMc2xHRwcTJkzoLdu4cSMTJ04cdL+6y+vU1dXF8OHDe8t2797NiBEjSlInm/PrxMfvKR8uvvjiPvc92SHaErOlrq4OgF27djFjxoxeox+Wl4Jzzjmnz/04uEiD7Iq6ujpaWlpyylpaWkp2jidOnJjj4584caK9kELdgKxtQ9HV8453vCOxLlOnTu1192zdurXXzTN16tSS1qnU5zerdaqqqtLly5drVVWVtZx58+ZpTU2NLly4ULu6unThwoVaU1PT5wLccfRKSlNTU687IpoupKmpKbYMFykSVN1dp4E4x0mJ1mFQ0jKXchuKhj+NLqoHjX+42Rp9V/pk6fxWap2iL4zwRZKGNIZf9aDxDzcbo6/qJimaqtvnad68eVpXV6eA1tXVlcToq9rXyRv+GCfy9ttvT/0QT5o0STds2KCTJk1KrEt1dbWuXLlSq6urE9+oLqhEI5mlOo0YMaLgPTNixIjYMkJjNHbsWH388cd17NixvcYpKaW636K/nzYNcign/5hSPk8uAHTVqlU5ZatWrfKGv1QPcfgA5m82D6DLFooLsnR+syTH5XUKjX+42Rj9UJexY8fmlIXGPymlNowD0eL/2c9+VvLnyQVgF8VVzPD7wV1H7N2716q8L6Ixulnh0ksvLbUKADQ2NtLW1kZjY2NqWUcddZQDjdKxa9eunAcyHJy1YeXKlX3uxyW6fmvitVwdMHv2bG6++WYuv/xyXnjhBS6//HJuvvlmZs+ebSXnTW96U+//b3/72wuWlyM9PT0MGzaM1atXM2zYMHp6euyFFHobZG0rhxZ/oWNKpYsrsnh+syDH5XVyUR/XLf60uDgvLtIgu5STJWzOL77FP/BMnDgxJwTNNsyqWHhYkrCxaLhXKVtvnuIUuyY216quro4tW7bkXOctW7aU7J5xUSeAhx9+2PiiMY3Thx9+2FoXcJNO2RVLlixhypQpVFdXM2XKFJYsWZJITn5q6SSppr3hd0hHR0dOJsCODrskpq7cRdGH7NZbby1YbssNN9yQ+FhP/9x2222JjnvFK15hVV6M6L3x0Y9+tGC5DWlccvX19ezduzcn1cLevXupr69PpEsWWLJkCddeey2LFi1iz549LFq0iGuvvdba+EeT8rW2ttLQ0NCbr8mKQt2ArG3lFNUD6A033JBKl/7KBluOq/ObVEbW5LjWpb+y/mRkJT1HKCd/gDocwLaRkTX3VVpcrrnb0NCQU9bQ0FD03OCjegqTNYNy5pln5pSdeeaZiXS59dZbc8puvfXWkhrJNJOUQjk2YWz96ZOle6a/sv5kbN26Nads69atJX+x9lfWn4zHH388p+zxxx8va8NfVVWl3d3dOWXd3d3Wcy4AbW1tzSlrbW0dmoa/v7K+cCFD1U0ImitdXMsB9O67705lJF3o4iIZWfjb+XHvpTT8aV6KkL0Wv4s6+RZ/YcC3+HNuzve///2pW28XXXRR4hvexTTz6G+HLf20Bjts6aeVkzQOOnpMS0tLal1qa2t11apVvUY/TZ1mz56duk4ujGQaOQORnuNd73pX4jrV1NQUrE9NTU1sGQMxKa3ULF68WCdPnqwrVqzQ7u5uXbFihU6ePNl6wfXwRdrQ0KCtra29Rr9Yz6EiDX9QsVQPnysZqm5Cx1zpMlDnpVRGspgsFzLKvU5ZSc8Rkm/8bYx+SP6EyHI2+iGLFy/W5uZmraqq0ubmZmujH2JzjSrS8Lto7VRi68IV0Rvra1/7WmojGSepVDFct2zT9ByidbjllltKXicXuDT8noHD9p6pSMMP6OjRo3Na2aNHj7Z+iGtra3NkpF3IIgsMVO+jVK3j8GaPUgmLy7iokwu84S8PbO+ZijX8kG6FHRcysoartLahnPzNdtzCVdRJVqJXXBp+F3VygTf85YHtPVOxhn/SpEk5ZWGGw7gAWl9fn1NWX19f1je8qyRXriKVXLXUXcnJ1z/sGdnIyP++b/F7BgPf4teDN2vUpVGq1psrXAxqAXrJJZfk5A+/5JJLStI6zh94DDdbP3YlLi7j6ty4wOVz4GKwOSv5713iok7ex6/Zeohd4GqgOXwJRlcMsm3VhnLCLW0cf1YMSpbumUo0/C5e0Fla8coVLutk8xxUpOGPtvCjm40rIpSR7w+3dYuopl8JydXD56pOWTKSrshSnUIDGaXcXT0u6lRXV6cLFy7MKVu4cGFZR9qVqk4VafhdPcRveMMbctxFb3jDG6xveNuJFX3VJ19GkofPRZ1cGsm0dXJF1gx/pQ3uuqgToF1dXTllXV1dJbtnXFCqOhUz/BWRnTO6eIQtdXV1nHvuuTmpW88991zrtLYHDhygqqqKrq4uZsyYQVdXF1VVVRw4cMBap507d3LSSSexc+dO62PBXZ1CojdMUtLWCeDss8/OSXt99tlnJ5aVFS6++OI+9webNM9SSNo61dXV0dLSklPW0tKS+P7NAi7r5OQ5KPQ2yNrWX4t/1qxZunnzZp01a5Z1K8WV7y38XRehpWnyyAxEnSC9jz9tqgVXIaou9Imeh2uuuSZx6zirE7iuvPLKktbJ+/iLY/sc4F09xckfK0ji3wezSHqUcLF0Gxku6qPqJoLAhT4ur5HrdVhLXSdVNwPWLshanXxUT2FsnwNv+IsQ5hVpbGzUtrY2bWxsVLDPLxL+bhpfdv7gcLjZDhK7IktGEtBx48blHB/uJ61T0nQLLo3k+PHjc44fP368tQwXuKyTZ+AAtLOzM6ess7Oz6HUqZvgrwsefhn379gGwY8cOpk2bxo4dO3LK4xL6RHft2sWMGTN6F8228ZW+9KUvtSofamzevJlZs2bl/E3DJZdc4kizZEyYMIFNmzbl1GnTpk1MmDChpHpVEpW2BKmIcPXVV+eUXX311db1qgjDP3fuXDo7O5k7d25iGdXV1axcudJ+CbOA2bNnF9QlLI9DaASib+bQGJSKGTNm5Pe+SiIjZPr06QwfPpzp06cnlgHQ3NxMR0cHzc3NiWWICPfee29iYxJe79WrV3P00UezevXqkl9vgLvvvrukv++K8LrkP9vlbPxnz57NzTffzOWXX84LL7zA5Zdfzs0332xlZ4Dyd/VUVVXlhC2G7pK4UKA7W6gsDmkTowG6efPmnLLNmzeXrLsNhw4ahYNLgykjlHPaaaflnN/TTjstlasnfxtMGaGcrFzvUP80s+CzBqQfd8siNnaGrPj4gfHAQ8BjwB+BK/o7plxm7row/K50aWpqyjm+qanJWkaWzi+ghx12WM7x4X4518mFHBd53l3eey7y8buQAeh1112Xc26uu+66RHVyMTDrKh+/zbOdJcN/NHBS8P9I4E/A8X0dU8zwu1jtx9UN73oFrosvvjixLuGN0dzcrB0dHdrc3JzI+GfJSLoehM9CnaLHJV2rwNXKTlm6Tq6vdfTcJKmTi1BMV9fJ9tnOjOE/RAG4C5jd13f6avGHmTTDzTazZnhcdXW1rly5srcraHtzuFxz14VByV/LM7xBbOVkyUimvdZRfforiyMj/+Erxfl1uZarq+vU2NiYUxYa7sGUEcpx8Wy7SLfg8jrZPNuZNPzAJGAjcHiBzz4CrAXWTpgwoehJcDE93NUNbxNmVUzGJz7xiZyyT3ziE4l06ejoyCnr6OhIJKe1tTWnrLW11bpOq1atyilbtWpVSYykKzlZ0qWqqkqHDRuWc+ywYcNKlicK0La2tpyytrY26/ObVoaqu/BoSJ9uoaqqSru7u3PKuru7E+li82xnzvADDUAr8M7+vttXiz9tQihIn9Nf1W2Lv7+yOHJctfgbGhpyysL5CTYyamtrc8qSrHAWfWiXL1+e81AnkZNmdrRrXdLIcVGfqJz+yuLIyUqL31Ur27f43Rn9WuA+4JNxvj+Qa+6OGDGi1/hv2LCh1+iPGDEitgxV9z7+sKWf5OFz5eN3mXgubcqG8Jj8a52mlX3jjTembvF/9KMfTd3iT+OKCK9H1HdsG9mWX6ek6yurZsvH78qv7n38boy+AN8HvhL3mGKGX9XN9PDQ+IebrdEPGah1bpPgIqpHNX2qaVU3dQJ0woQJOTLC/bS6pDH8aerlSpcbb7wxJ1okfKHZ4urey0pUj6q7SBof1ZPe8L8uULgN+H2wndPXMX0Zfs/QoNBLJ2nLduzYsTlloXvERoar3kfaOrmoTygn3JIm5PNkj2KGf9Bn7qrqKlUVVZ2mqicE2z2DrYcnHqNGjcqZ7j5q1KiS6XLgwAGqq6t58MEHqa6uTpTyuq6uji1btnDUUUfxxBNPcNRRR7Fly5ZE6XGXLVvGmDFjWLZsmfWxIWnr5LI+Ieedd17iYz3lQUWkbPAMDKNGjWL79u056Q22b99eEuNvGi/GUL7xjW/sNZBheVz27NnTayyPO+64XiO5Z88e5zr3h4s6ua5PbW0tq1atora2NtHxnvLAG35PUUKj/+ijjzJhwgQeffTRXuNfCgp1WZOwZ8+eHBlJjWQ0uVpSXNTJVX0Auru7OfXUU+nu7k4sw5N9yt7wz58/n/r6ekSE+vp65s+fX2qVUpEl1wqYLKVRfWyzlmYRV/fMmjVrGDduHGvWrEmsS9aut4sslpX2TFYkhVocWduKDe5W2ko9rsIwXYGjQcws4eKeATdRPVm63q7qVGnPZLlDVqJ6kmzFDH+pVq4fKEIjECXJxCuX+hQyTOVs+F3cM+E5yI8ztz0vWbreof5p51tU2jNZ7hQz/KIJ/aSDycyZM3Xt2rWHlIsIXV1dDB8+vLds9+7djBgxIrH/t5SICB0dHTkLcWzcuJGJEyeWpD4iwpFHHklnZ2dvWbhfjucX3NwzfblBbM5Llq63yzpV0jNZ7ohIq6rOzC8vax+/y5Xrs8Lxxx9PVVUVIkJVVRXHH398SfUJVxIrth+HLPmx6+rqOPHEE3PO8Yknnliye+acc87pc7/cqMRnsiIp1A3I2jZUfPyupqq7wkXKhiz5sVUPrnE7a9Ys3bx5s86aNUvBbq1bIv7v2267reJ8/EnXIlatvGey3KESffyqbqZSu0j74AIRKZhp0SbRm0vCcxrVJ9y3kZEVP7aqOcfNzc0590xzc7P1OY6ekyQGMsRVao20FKpP0nq5eCY9bqhYw58WF4neXAHpUzu71idtemcXMlyStXOcFbJ2nTxuKGb4y9rH74L29nbmzJnDXXfdxejRo7nrrruYM2cO7e3tg66LiHD11VfnlF199dUlXRzahQ86S37sLJ7jrJCl6+QZYAq9DbK2DWSLH9Iv5uIKF6mdXeLCB50lP7Zq9s5xVsjadfK4Ae/qKQykX8zFJS5SO7vEhQ86K37skKyd46yQtevkSU/FGv7ojRpuNuQP7IZbEh+/ixziaesT4mqAzYU+rurkat2ELNXJhZwsnRdXcrJ0/6q6y6XvgrLMx59k62vpxXBLGoJW6MZIcoO4CMWM/vbFF1+cWBdXIXXR3//c5z6X6vymXfDa1Upp0Tq0tLSkvmfOOuusxNcpPCbNbNmBOC8XXHBB6jqlWU7S9f2b9t5ztXqWC8p2Ba4kW3+Gv7+yvnAhIzzGxVqjLnRxNW3e1fmtrq7OKQsfQFtdXKyNnLV7Ju16xFk6L+ExaReXcXn/urj3XK2X64LQ6EcpmzV3bbe+DP8tt9ySUxa2/OMC6O23355Tdvvttye64dva2nLK2trarHW5+OKLc8rClr+tLl1dXTllXV1dieR87nOfyykLW/42MlauXJlTtnLlykS6bNiwIadsw4YNieS0tLTklIUtfxsZZ511Vk5Z2PK31WXVqlU5ZatWrbLWxdV5ueCCC3LKwpa/rZzly5fnlC1fvty6Tq7uXxf3XlVVlXZ3d+eUdXd3J1qGNC1gF3ZbsYY/S6033+IvLMO3+IvL8S3+Q/Et/uL4Fr8evDHDln90Py7RY8KWfpIb3vv4+5bhffzF5Xgffy7ex18c7+M/WLFDNltcyFD1UT0DKUM1W9ErrurkQk6WzosrOVm6f1UrL6qnrNMyezwej6c4FZmW2ePxeDz2eMPv8Xg8Qwxv+D0ej2eI4Q2/x+PxDDG84fd4PJ4hRllE9YjIVqCjn6+NBral/CkXMipRF1dysqSLKzlel4GVkyVdXMkZTF0mquqYQ0oLxXiW40aReNXBllGJuvg6eV18nSqrTt7V4/F4PEMMb/g9Ho9niFFJhv+bGZHhSk6WdHElJ0u6uJLjdRlYOVnSxZWckutSFoO7Ho/H43FHJbX4PR6PxxMDb/g9Ho9niOENv2dAEJEqEXlPqfXweDyH4g2/I0SkutQ6hIjIW0WkpNdWVQ8An3IpU0SGu5SXFZLWSwzjXevjcUvQCJrlSJaTZ7tsDX9w039ARP4z2J8gIq9OKKtaRMYFMiaIyIQEYv4sIv8tIscn0SHQ459EZKmIPCYifwm3BKLeG+jzXyLyqqT6OGC5iFwpIuNFpCncbIWIyCwReQx4ItifLiJfTyBnjIhcIyLfFJHvhFsCOfNFpNH2uAJyUtVLTWTGPQ70qBaRh1LK+LmILCu2Wco6VUQeEJE/Bc/A00meg7TPtYj8WyEjKyKjROTbceUEjaCv2fx2Hzh5tmscKVMKvg4cAN4AfBbYCdwJnGwjRETmA9cDWwJ5YFa2mWapz3Tgn4FbgpvlO8DtqvqihYzvBrr8D3AmcBEJXs6q+gERORy4ALhVRDSQvURVd/Z1rIisUtXXichOzHno/ciI1sMtVHlv8PejUfWAl1nIAHM+zgaWYZT4g4i83lIGwF3A/wHLgf0Jjg8ZC/xORNZhrvN9miw8zkW91onIyar6uwS/T/C7+0XkgIgcoaovJBRzY9LfL8C3gU8ArSS8To6e61dizu9HVXV1IPdyTE/2K5YqPSgi7wJ+kvBeAdI92/mCynID1gV/10fK/pBAzgZglGPdTgf+BnQB3wOOjXlca/C3Pb8soR6jgI8DzwC/BP4MzC/1tUtQj984uta/d6iTYIz27cE9tAB4+WDXC9Nb2Ac8BbQB7UBbgvrcBWzEGN2bwq2U1zulDCfPNTALWA/8APgdsBg4OoGcnZgXUDfwYrD/Ygq9Uj3b5dzi7wn86uYpFBnDwTe7DZuApK2cXgJdzsW00icBC4EfAqdhuuOviCFmb9Bb+LOIzMO8PBoS6PI24MPAscD3gVer6t8DX/JjwKI+jn2nqv4k+L9RVXfY/n5E1nDgk8AEVf2IiPwT8EpVvdtS1KbAR6oiUgtcATyeQKW7ReQcVU3tHlFVFZHngOcwhrcRWCoiD6hq3LENF/U62/L7xfhJsKUiuMZfBI4H6sNyVbXp5T0kIv8d6LM3ImOdhQwnzzXwKPBb4M2Y3ve/qeqztkJUdaQDXVI92/kKleUGvB/TRf4r8AXgSeDdCeR8G1gFXI0xUp8EPplAzl8CWbMKfBar5YRxUzUAx2C6bz8BXptAl1uB1xf57Kx+jl1X6P+E1+hHmG7xo8H+cBK0ujFZCH+I6bb/HbiNBK05Dra69gT/J2p1YQx0K3Af8G6gNiivAp4a7HoFsl4CTAi3hDKGAVOCrTahjFXAWZjex0TgBuCzljIeKrCtsJSR+rkGPhg811dh3OIzgEcwBvclMWW8Kvh7UqEtwflN/GxHt7KeuRsMbpyF6XY/qKrWrUARub5Quap+xlJOg6rusv191wQ9j+WqembC49er6on5/yeUtVZVZ+bJ/IOqTk8qMwuIyGeA76jqIanCReS4JPdhCl3mYHqX4zAvj4nA46rabCnnDIxb8hnM8zQeuFBVf2Upp1VVZ4hIu6pOjZbFPL4KOF9Vf2zzuwXkpH6uReQu4GPR6ywiAlwG/LvG6MWIyDfV9HYLDZ6rqr7BQp9Uz3b+L5fdBlQDTziWOTzl8f8FHA7UAg8CW4EPWMqYCfwUWIdpMbWRzF/7IHBEwno8AZyIad08HvyfqIUCrAEO4+B4zMuB3ybQaTLwZUwPaFm4JazfHMxA5I3AeZbHNvW1JdDle8CRkf1GzAvFRsYfMP7e9cH+mcC3E+jSinHDhfuvIMH4UnDNq4JrNQ94B/CkpQwnaYsDWQ1Agyt5EbmJelUOfjfxsx3dytLHryYK4UkRmaCqG9PIEpFTMN3CBmCCiEwH/lVVL7cU9SZV/ZSIvAPTanon8CtM9z0uPwT+HTNAl2S8ImQX0C4iD2AGmAFQ1Y/FOPZZjJEF47/+cuQzxURRxeV64F5gvIj8EDgV45+05WeYa/RzUpwXEfl/GHfaD4OiK0TkVFW9OqaIVsw5EIxLZUfw/5GYgdHJlipNU9XOcEdVd4iIbQ+rR1WfFxMrXqWqD4mIbcQJGNfOkxFd/hSMO9hyBcal9zHgc5gX0YWWMpaLyJUYV2H0/t0eV4CITMEMyDYF+9uAD6nqHy11ico8EngX8D7gOEwvK+6xtcBcIIzaWgl8Q1V7LNVI82z3UpaGP6AR+KOI/JbcEzDHUs5XcBMqGD4k5wJ3qOoLpldoxVZVtYp5LkLigTp10Y08KOuBIOTxtRgDeYWqJll5aI+q3uRApXOAE9TEVSMi38NEbMQy/Ko6OTjuW8BPNRgkFpG3AG9PoE9VdABdzBwH22eyU0QaMGGqPxSRvxN5HixYKyK3cLCh8n5gra0QPRhWugsT6JAEF2HA38T49B+CXlfWtzBROrERkcOAt2GM/YnASMy1tnKBATdjbEQ4T+ODQdkllnLcDMIH3YeyQ0ROL1Suqg9byvmNqr4mrR86aE2+HfgH8GpMK/BuVX2NhYyzMPG5D5IbzWB9oYMbdkK0FZeU0E9p8f2T+vpc7aIzEJH3Af8E3E/yKA9EpA04I2w5BoZ2papazdmI+q/7Kosh50PANcAdmBfj+cAXVPUHMY79OMat6LeYuQAAGMNJREFU8hiwG+NeeT9wBPBDVX3eUpc6jKF9XVD0f8DXVXVv8aMKynkAE2TRGew3YuazuIo+iqvHIc+w7XMtIosxUXn3Y8J2VwAbwgbAYOvjkrJt8dsa+D5wEiqoqp8Wkf8CXghcUV2YloINFwGvwrQMopNOrAy/iLwV48MeBkwWkRMwkRW2vaGQmZbfX9jHZ7buIoCpmBbSG8g9L7ZyvgisDwbaBNPt/rSlDIDNInIdua3jzbZCVPX7ItKKcYcAvFNVH4t5+DGY3uqrMK7B1ZgXwc9tXCLQO2j4HVV9P7muvSSMLuC+eomlPi7CgP8iIv+BcfcAfAAToWPD8Rh33uOYAfP9wYSpJOwXkZer6lMAIvIyEkxOcxQuW9Yt/ujM0mEYY9mldjNLEZHRwFeBN2KMwf0Yl4RViymQNYVDL8j3LY5/UlVfafu7BeS0Yoziykgv5lFVnZJQ3r2q+ua0eiVFRDYAx6tqtwNZR3NwdvdvVfW5BDKaMOMXr8fcg7/CvFitDG4gqxozE7i3EWYzbiUiwzAv5lnAKcHWqapWqUNEZBXwhrTnOLj33hHWQUQmYtxiffYC82T8CDOe8iFVnRK8CNao6gkWMhqBz5Dbg7lBLeelBJGDF2DcT9sws3mnqOoWSzlnYUK0/4KxMxOBi0JXlIWcVRyc3f9Wgtn9qvqfNnLKucXfOyEiCLF6G8aXHJvgoftg0NJJRRA+dgbG8N8DvAUTRxzb8ANrROR4i1ZfMXoKjDEkHhS1Nfoi8s5+5Nm6rh7FuM7+bnlcqM+rVPWJiAvqr8HfcSIyztZlFBj4K0RkhKom8aeHekXTCuwnSIuBXVqBwzDRZEcE22ZMD8CWvwCrxeTViY6Z2fYArgVWicjDmPqcBsR2Ewa8XFXfKyIXBDrsFssBs8DAWw14FpHzBOYaXS8iMzC+/t+JyF9VNfZ4gao+GPZcgqInbd1oAYcFskRNmOkNwct2aBj+KGq6LT8LjG/srnvQdXsf5u2ZlvMx+XrWq+pFIjIWu4geMC+u34vI0xhfdpgfxzZv0B+DelUHN9vHMG6AfhGRr6jqx0Xk5+Tm6gFiD56/tY/PrF1XGKP/hIj8jlwff1zX1ScxxqeQC8raZRS4Bm8hfSTYFRgXRpLe5TeBZswktN9gru+XbVu0EZ4KtirMAGYiVPXe4AUbNsI+nmBAvzsYowpn5b+cyHXvC0f3b0FUtRVoDSKOToupT7FAkdeICGo5TwJHs/vL1vDntSqrMN3dPQlErRKR/+XQ0DGrViDwD1U9ICL7xCRR+jtmEowNrtwp8zEtr72Y3CL3YULr4hD6RBMn3VLVpNEcxSg4GScukYHpt6hqzj0iIvUFDukPV0nj0qQVmADUYXK0/A3Ti+ns84giBD3fkap6ZUJdCvWqwjGPCWLCrm2epxs4NAw47j2V+v4NEZH+IsniGO1/L1AW9urGY+Yk2VAoXPZDljLK1/CT26rch4mdtx1MBQj9hp+NlCUZOFwrJs73Wxj/5C7g1zYCgq4bwWBYEoMUcq6qXosx/gQy342JHulPh9bg37UEL7Pg+GqMoekXEfmAqt4mIp8s8htW7gOHA/lrMBPR+ivrF1XdlOd9SJJF8i/AShH5Bbk9mX7Pj6q+OXB/NGP8+/8GTBGR7cCvVTX2yzLo+Z5qrX0uznpVqnp/4L6wDgOO3L8nqOpXo5+JyBWAzb10GcbN+GPMi8w6PltVc3q/wXm+DjNHZr6tPGBSEDLbGy4bPNu/sRFSzob/Fg1SpYYEJ9XKD6yO4tYj3fwWEbkXOFxV22xkSJHp95iH24arOdTIFyrriwcxA95hGorDMAPfcfyaI4K/qRJTiaMU0SJyFPBS4DAxE6TCB/hwTOvJFldJ4zYG27BgsyJwcT4qIp2YnsMLwHmYcGLbXtLvA//+HeT2fGO55cJelYvnSUQeVNWzgF8UKIvLhZigjSgfLlDWF0djcjG9F9O4/BGwNBq1FJdgcPc/MPfxAlV9wFZGgItnu6wN/yIObakVKitIsdZoiG2rNHpjquoz+WUx+RymlbNcVU8UkTMxYWhxdXgLZpLSS/O6qYdjblwb6jWSe0hVd0nMlaJU9RtBD+FFVU08fqKqrwv+ps1seDbmoT8G82INDf+LmDh6Wy7DGJCXYtws95M72SgWapkPKoqIfAzzEp4F9GB6Lmsw6wMkGdytB54nt2WeJJT43cC9qrpTTMjrScDnVHV9jGPrMS/i0UFUTvQF/dKYv38BZgB2suQuADMSsIq6CsZeWjCNuWMw6208JiJXaYy5FoE+52J63i8A16nqKhsdInJcPtvlZ/jFpFiYBYzJM96HY+cvC43JKzHhfeFN8lZMGta4+qS+WSOknX6/GeOimYNxN4XsxCxsYUOXiJwU+mZFZCZmclosAvfBBaQcOA9eIH9U1cSrDanq94Dvici7VPXONPoE8rZhYvdTISaV+KcwPbpoCHAct8gkTCvvE5ogTXA+Dsdl/kNV7xCR12F6jP+NMZ5xJjL+KybH/DjM/Rt9Qf9vzN9fg0k7Mppct9NOTO4ra4JxiwuA2Zjc9619H5HDzzHjL88DnxKRnJTdFoPNLp/t8jP8mC5xA0b3aEvwRUxkTSzC1paI/AqTfGxnsH8DkS5mDFzcrCHh9PtfkWD6var+AfiDiCzWIAdI8DIanyDa4+PAHSISDtIdzcGp9HFZnXbgXB3mZQJmBL2w6KzSf1PV62yEiMhkjH92Ernx97YRIz/EnJvzML2ICzHJ/fpFVfvssdoStGgXYQZSwcS9X6Gqfy1+VEHCsY5zgW+q6i9E5PNxDgx88l8VkfmqGi+v/KEyOoAOzHyGVIjIZzH1eBwzc/dqVbVtXbtyJR/ybKehnCdwTdQCaXETyHkSkyxrb7Bfh8mIaTWRKs3NGpExAtOqTjv9fiWmZVCDeRn9HTMBpt+WgYicDGxS1ecC//W/YhLOPQb8p9olykqdijaQ8ytMnpRUeZmkQJppEVmnFpOLgmP+gEkal5NMz3YQWg6mMG4LQ3ZF5HeqarV8qAvEpFpYTO5M1/er6mxLOXdj3F+zMW6ef2AmytmkSijkLvq8TYNBRF6LeZEdh2ksVmM5wVNEDgBPY1JiwMFxpqRh1qHck2zqknfsqZiop4mY5zvUxWrmbjm2+EPqxMQyTyK31WUbjfN94Lci8tNg/+2YxQ6sUNVFwYBfvj6xJnAFLo27g8GxA5iUvUk5QlVfFJFLgO+r6vVi8tTE4RuYLjqYVtM1mNbtCZjEVza9qlStHRE5FjOr9T/yPjoN0523pVpE6iIv+cOIGamUh6ukcWHL7dnAF7yZIJtkCRijqt+N7N8qJh+QLe/BhCXfqKqdYmZKFwpp7ItC7qKbiecuCvlfjE/+Dkyo94eItwpeFOucPDG5hQSRZAGp1yOG8jb8d2B8h7eQ4gSo6hdE5JccnJBxUZyBqHxE5AeYfPO/j+ijxJy5q24WvA6pCR649xAJ6YxJdaRV/15Md/1O4E4R+b2NoKD39C4OfRl+ttgxeXwF073OGawMQhYXYB4CG36IWfT6u5iW0odJ9oL9qpjJgqmSxgGfF5EjMKGYizDjQtb+Wkc8LyIfAJYE+xdg/NK2jCbI6ikiE4KyJyxlJHYXRVHVDSJSrar7ge+KSOxMrMHxqT0KRbAOC43wgqr+Mq0C5Wz496nqzUkPFpNvJeSZYOv9zMalETATk08mje/MSa5tzJyE+4BVqvo7MQmh/hzz2GoRqQl8mWeRO93e9n65CxPN0ErMmZd5jM03+gCq2i4ik2yFqeqXAjfNGzEv5fswXWZbnCSN04NJx17AkS84Bf+Cefn8D6Yua0iWVvkXHFyzoB7Tan4Su5Dkv4nINzDuoi8FDYgqSz12i8lj9HsxyROftZUhh4YR936ERThxARJHc+FmPeKy9vHfgPFd/5TcExDLYItJixDeoHCo/84u253IHZhl2hJHWIhIwQUrgqgUGzmjbMcFIsdeiwkb24aZHXqSqmrgdvmeqsae6CMpEsMFx/9ZVf+pyGcbVPXYBDJPxIT7vRvjv71TVa0G4cVR0jiHg8SZJYiIuVxVY+edFxM2/GagXVX/HPRep6rq/RYyJmLsQy2mF3UEJs30BqsKOEREpnHotbYNl3UzblbGhv/pAsXWBtsVwQU5ATMAmSSfTChnTHBcrOiOIjL+jHE5fRf4pW0vJBgYOxq4X4MkZCLyCswSdjYDbN8EFhVqtcc8fglmke1v5ZVfAsxW1VhRRoHuFwTbNkwkzZWqmqS1j4j8DPiIqiZKGheR42SQOKUOiyjcqg11SZ3oTJKtVfA64J9U9bvBM9GgqoWe+bJARL6DSdPwRyK9RFX9l5LoU66GPy1yaG6RHKy7TikWhhERwcy0nIfpjgpmUsYiC394vrw3YrrvJ2OmnN+qqn+ylZUEEXkUc3PXYBZQ+QsJks6JSXT3U6Cbg7HLMzFRGu/QmCmVg+iM/wMuDlt8IvKXpI2EIGpqGpA0aVwo5zdqsVDPQJDXy/wMeTN+E/Q2o2GmVZhBzFFqsRBLMH4yE5PA7hUiMg6zqp1Nb/M8zITI/OiXpO6ZVIjIY2qZKruInLGY8a1xqvoWETkeOEVVrca7ytbwS8rFGiRYVcpV1ykNwcPyFkwr8umg7GWYSIZ7NcXsVzGzf2/DpFH4A/BpVbXKIZTgN3dwMAfSIdgOmgV1CF1Gf1TVFZbHvx0T4XEqJvnX7ZiUH4miNtK85PPkOFlZzBWFwl0TyIi+OMIcWndqXnK8fmT8HhO+u04PrifRFrfBEHx/AyYMuT3luJsTROTbwEJNmXI9CET5LnCtqk4XkRpMRmCrHhVagpXiXWyY7vqngEeD/eHA70uoz2sxLcBdmBbqfkzKgjjHrsesXJRfPia4qLa6jMLkj1mLGWx7J6bVMxN4ehDOxbpS3x9F9BqB8e//HDN4fjPwpoSyxmImXp0HvCShjC9iZnU+DDwUbCtKeH4ycd0wcf+9+gTXrc1SxkOYBUpKXp9An9Mxg/hPYmYQt9vWKZDzu+Dv+kiZtd0r56ie1Is1hKSJv4+QJm64VgtkH1TVrWImUdnya8xEnLdr7szLtSLSkkCeLS+RPnIhqf3iHk5QM16xGFgsZtbuu4GrMC3u2IjIezCx5SsxLoRFIvLvqrrUUqV3Ay9TByuLlRrJzYtzCGrnBvtxENVzpIhcinFZfqufY/L5FHCPmAVhrDKfDhDfxkSC5YznJKBLREZxcK2C15IgtXc5G/7EizVESRt/H0WTxw339eAnMQqv1KApUEDHLyWQZ0s1Jq1GmnjlAUVNCotvBpst1wInazC4Gww+LgdsDX+qlcVckBeyOFxEXgw/ws4nfgpmfYElmBTB1tdeDi4g/xVMeOuLmFxa/6n22Sy/gOl915Mg8+kAsFVV+3w5xuSTmLxiLxeR1RivQOxJlSHlbPhv4NDFGj6cQI6L+HtIFzc8PfLARQljoWMRbXUV6vxYtrrS8KwmGJQuI6o0N6LneezjzCH9ymKp0fSZT0OOwsTdh9kxfwEsUdU/WsgotoC8TVK0kHGaIpR4AFgvIosxbsbotbYK51TVdcEY0ysx9uFJTZC7p2wHd8HEq3NwsYZHCrlLYshIHX8fyJmIWTt1GCWKGxaRrfTR6tJBChN0MUiYZYIJNNM4OMv1vRh/7VWWcpwMEmeNYMLVBRh32GfUfp5E6gXkg8bXcrWI/R9IxMwWz0fVMpxTTGqXcznUNW2XRr5cDb+YNTUXA8s0wYLXcnBNzpGkiL8XN1kjnRDcFGGraxrJWl0u9Egy8znzBJPYxqrqajFLf74u+KgTk0zvqQQyx2JCbsEMapbM7ZOWwOCfi7n/JmFcEt9R1b9ZyjkCY+xPDf4eiYnOiT2TOHBhjcA80z2UOJzTFSJyD2aJ2fy5H1azgcvZ8J+OaWmdi4mmuR2T5CxW2FgwaDQWE98d5TSMqyJWXKxEsjuKyJ2q+q6YVRhQ0ra6PIciJvNkodxBUzGrKvW1yHwhefmDxKcBSQaJS46IfB8TcnsPcLuqPppARv4C8o9gevJJF5AvOa4nyNmGtRajbH38QXf44aCV+wbgUszqQ3Hf6G/DTQKwqDulJLOGoxRodd2EmQTlSY/T3EG4GyTOAh/AhMheAXwsMsZk09JOvYC864mZ/7+9+3+xrK7jOP58jSzulsS2qBVWgkFZSSsuJjgQa0v6y4qkllpRUJDS/pDSEgQW9g9EJKxoQmsURt/8RmgrsrSaxLpj6tbWFLRB/WJKXyDLWrZXP3zOHe/cmbs7594zc++55/WAhXvOnfPZz+zd857PfM7n83434EjD7T0q6cpxp7BaG/iBXlrdqykj/0uol2mxqZvYQ15vuIFR11dHGXXFKW09xXtbRmivqYfEE2d77H67mQLyjRV9b4Jr7nxeg18AD0iaY4wprDZP9XyfUlT6Mcpmrp/ZXvP6WDWUAEzSScpIR5Sbv1e0YcPnFKvUBL3nHSMXJ4/VqaHcQX3XrfaQ+KjtLw6/qhtUKoLNU34A7KakfTjVD96pJOnrtm/te6a4TN0VXCo5yq5hzB3JbQ78V1Ge2o+Ui7/pmzhmnxrKHTTQZv9D4idtd3ZaTsMLyD9NCXS1Nj41tDFzLJJ22F5oMM3HIWBn3X+LFe20LfBXN8pQa10Xux43cXTDuLmDTtHuHHCT7e820V7bSPoa1dr9BpZXr7oxs+7D1HE1vepP0n7Ks8RHGWNHchsDf2897LmUkUHvpruC8h9md8321uUmjhhG0huAPcB5lCWPj1fHe4HnbV8zwe7NBEm/oZmNmeP2o9FVf1qeBG9Jl5ZzPg58sjcyUCnWsN810r9GTIKkh4C/UXIq7aIMYgR83nat8paxuqY2ZjbQj6XNjNO0sbHNq3reOvChvshoZfQiNtoFrtLoSrqXkt7j7WvdgxJrcjZwTFL/xkxP4LepRlf9qaSRX+0hca3VSm0O/E9I+inLV0RMxfbsiNNYyq1i+6SkPyfoN+6Ovte9zXE3TqAfvTxcAraMkQSvZ2/f683AdZS6B7W0dqoHQNKHgQ9Uh38F3mx7zwS7FHFafUuAYfky4Cy7bZBW1lf+se07J9ur5kk6bPv9da5p84gfSnWfy+krnD3R3kSsge0zJt2HWaXV6yvL9hUT7VhDJG3rO5wDdlASQtbSusA/6x9sRIzlt5T8W7v9Wn3l2ybbpUYtUOb4e3W5jwOfqdtI6wI/s//BRsTorqXM5R+U1KuvPLUFgeryiHWiB7UxL8i1lFUQByV9U9IuZuiDjYjR2X7Q9o2UYi4HgVsppUDvknTlZHs3Pkl7JG3tO36jpM/VbqetD3clvZ6Ss+ImSuKlbwMPjJu1LiJmi16rr3yD7V2T7s84JD1n++KBc7X3B7Q28PebpQ82ImIYSUeB9/V2JFdp6V+w/d5a7cxC4I+I6IIqo+v5wN3VqZuBP9n+Qq12EvgjItqhSuR3MyXVB5Q8T/fWzVKcwB8R0SJVMfp3UZZ1Lto+cZpLVraRwB8R0Q6SdlIqDf6RsprxbcCnbB+q1U4Cf0REO0haAD5me7E6fidwv+0dddpp4zr+iIiu2tQL+gC2fwdsqttIG3fuRkR01ZEqlfd3quOPA0fqNpKpnoiIlpB0JqVa21KdZmCf7f8Mv2qVdhL4IyLaQ9I5ALZfGrWNzPFHREw5FXdIehlYBBYlvSTpK6O0l8AfETH9bgPmgUttb7O9DbgMmB8lO3GmeiIippykXwIfsv3ywPlzgAN1k7RlxB8RMf02DQZ9WJrnr72cM4E/ImL6/XfE91aVqZ6IiCkn6STwympvAZtt1xr1J/BHRHRMpnoiIjomgT8iomMS+CMiOiaBP6JhkpL8MKZaAn90nqQvS1qU9JSk+yXtlfQOSY9JWpD0pKQLq6/dL+kbkp6W9AdJ11fnd1Zf9zBwrDr3CUmHJT0n6W5JZ1R/9kv6laSjo+y6jBhXRibRaZIuBa4DtlM2wjwLLAD3ALfY/r2ky4B9wAery95CyY54IfAw8MPq/CXARbaPS3o3cAMwb/uEpH2UFLq/Bs6zfVH192/dgG8zYpkE/ui6eeAh268Cr0p6BNgMXA78QFLv687su+ZB2/8Djkl6U9/5w7aPV693ATuAZ6o2tgB/AR4BLpB0J/AT4MD6fFsRwyXwR6w0B/zd9sVD3u/Pfa6+168MnL/P9pcGL5a0HbgKuAX4KPDp8bobUU/m+KPrfg5cLWmzpLOA3cC/gOOSPgJLKXG312z3CeB6SedWbWyTdL6ks4E52z8CbqdMD0VsqIz4o9NsP1M9kH0BeBE4CvyDMh9/l6TbKXP/3wOer9HuseraA5LmgBOUykn/Br5VnQNY8RtBxHpLyoboPEln2f6npNcBh4DP2n520v2KWC8Z8UfAPZLeQ3moe1+Cfsy6jPgjIjomD3cjIjomgT8iomMS+CMiOiaBPyKiYxL4IyI6JoE/IqJj/g80dW0nFUDWjwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.boxplot(score_dict.values())\n",
    "plt.xticks(list(range(1,20)),score_dict.keys(), rotation=90)\n",
    "plt.title('Average rating of each genre')\n",
    "plt.xlabel('genres')\n",
    "plt.ylabel('average ratings')\n",
    "fig = plt.gcf()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Consisting with this graph, we could obtain some information about the train dataset we are going to deal with in the coming sections. First and foremost, we could see in the train dataframe, there is no significant variance in the average ratings between the genres within the train dataset, with all their average ratings concentrating around 7 with few spreading in the range between 7 and 8.\n",
    "\n",
    "Moreover, if we focus on the average rating of each genre, we could see that among the 2000 users we selected, the film noir tends to have the highest average ratings (about 8.4). Moreover, the horror films tend to have the lowest average ratings (about 6.5).\n",
    "\n",
    "What we could learn from this graph is that the average rating of all film genres do not vary a lot, and the negligible differences in the average rating may not greatly affect further prediction that takes the rating as a predictive factor.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##  Graph 3 Number of films in each genre"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "count = defaultdict(int)\n",
    "for index, row in train.iterrows():\n",
    "    genres = row[4].split('|')\n",
    "    for gen in genres:\n",
    "        count[gen] += 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.bar(count.keys(), count.values(),color='g')\n",
    "plt.xticks(rotation=90)\n",
    "plt.title('Number of films in each genre')\n",
    "plt.xlabel('genres')\n",
    "plt.ylabel('number of films')\n",
    "fig = plt.gcf()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "According to the graph above, we can see that in our train dataset, the distribution of the number of each genre of film, far from a prefect equal distribution, is indeed an uneven one: The drama, Comedy, Action and Adventure have the greatest number, while the documentary, Film Noir, IMAX and Western are the most numerically modest ones, and the difference between the maximum and minimum are more than 100000.\n",
    "\n",
    "This graph provides us with a number of hints for further explorations: since we consider the amount of the film in each genre as one of the predictive factors, we would definitely put this excessively various distribution into consideration, and hence may not use the absolute number of each genre yet the relative proportion of each genre. (for instance, if a user only watch a modest amount of documentary films, it may not indicate that he or she dislikes this genre but it is short in general proportion).\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Graph 4 Number of films each user rates"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "count_rating = defaultdict(int)\n",
    "for index, row in train.iterrows():\n",
    "    count_rating[row[0]] += 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(count_rating.values(),bins=500)\n",
    "plt.title('Number of films each user rates')\n",
    "plt.xlabel('number of films')\n",
    "plt.ylabel('number of users')\n",
    "fig = plt.gcf()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Considering the histogram above, we could see that far from a normal distribution, the number of film users have rated is an extremely positive skewed curve: while most users have only rate around 100 films, the minority rates more than 500 films and an outlier even have rated 2500 films.\n",
    "\n",
    "Therefore, we could see that the number of films each user has rated varies a lot and outliers do exist, and therefore in the further data exploration, we may further consider the proportion of films in each genre rather than just focus on the absolute number.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Graph 5 Variance of ratings of each genre of films"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "variance = defaultdict(float)\n",
    "for index, row in train.iterrows():\n",
    "    genres = row[4].split('|')\n",
    "    for gen in genres:\n",
    "        variance[gen] += row[5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.bar(variance.keys(), variance.values(),color='y')\n",
    "plt.xticks(rotation=90)\n",
    "plt.title('Variance of ratings of each genre of films')\n",
    "plt.xlabel('genres')\n",
    "plt.ylabel('variance of ratings')\n",
    "fig = plt.gcf()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Regarding the graph above, we could see that the difference between the user's rating and average rating varies a lot between genres. For the action films and adventure films, most users in the train dataset tend to give much higher scores (1.25 and 0.50 higher than the average ratings). Moreover, for the drama and crime films, most users tend to give lower ratings (-0.50 to -0.10). And the rest would have relatively small differences between general ratings and user's ratings.\n",
    "\n",
    "Therefore, from this distribution we could learn that, in our further predictive model, we shall consider these factors, since sometimes a higher rating on a certain genre of film does not indicate a favour but just an overall phenomenon.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data Analysis & Results"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In this part, we used three methods to suggest movies to users. The first is the cosine similarity of vectors of genres. We formed a vector of each user containing those genres that user loves most. When suggesting movie, we will compare the cosine similarity of the movie genre vector and the users' preferred genres vector and we will recommend the movie if the two vectors are very close. The second and third methods are classic methods. We find the average rating of each user on each genre and the number of movies each user watches on each genre and generated two dataframes reflecting the data. We predict whether we should recommend a movie based on each user's avarage rating on each genre of movie and how many movies in each genre they watched before, respectively."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Cosine Similarity"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "# we define a function to calculate the cosine similarity of two vectors\n",
    "def counter_cosine_similarity(c1, c2):\n",
    "    terms = set(c1).union(c2)\n",
    "    dotprod = sum(c1.get(k, 0) * c2.get(k, 0) for k in terms)\n",
    "    magA = math.sqrt(sum(c1.get(k, 0)**2 for k in terms))\n",
    "    magB = math.sqrt(sum(c2.get(k, 0)**2 for k in terms))\n",
    "    if magA * magB == 0:\n",
    "        return 0\n",
    "    return dotprod / (magA * magB)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "# collect the users preferred genres into vectors\n",
    "users = {}\n",
    "for index,row in train.iterrows():\n",
    "    if row[0] in users:\n",
    "        users[row[0]].extend(row[4].split('|'))\n",
    "    else:\n",
    "        users[row[0]] = row[4].split('|')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "# eliminate double collected genres\n",
    "for key in users:\n",
    "    users[key] = list(set(users[key]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "# the true label\n",
    "answer = list(test.apply(lambda row: 1 if row.rating >= 6.0 else -1, axis=1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "# the predicted label\n",
    "pred = []\n",
    "for index,row in test.iterrows():\n",
    "    if row[0] not in users:\n",
    "        pred.append(False)\n",
    "    else:\n",
    "        counterA = Counter(users[row[0]])\n",
    "        counterB = Counter(row[4].split('|'))\n",
    "        res = counter_cosine_similarity(counterA,counterB)\n",
    "        if res > 0.3:\n",
    "            pred.append(1)\n",
    "        else:\n",
    "            pred.append(-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.7698619171502902\n"
     ]
    }
   ],
   "source": [
    "# accuracy of our model\n",
    "acc = sum(np.array(pred) == np.array(answer)) / len(answer)\n",
    "print(acc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "# precision,recall and f1_score of our model\n",
    "precision,recall,f_score,support = precision_recall_fscore_support(answer,pred,warn_for=())\n",
    "pre = precision[2]\n",
    "rec = recall[2]\n",
    "f1 = f_score[2]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# we use boxplot to plot the accuracy, precision, recall and f1_score\n",
    "plt.boxplot([[acc],[pre],[rec],[f1]])\n",
    "plt.xticks(list(range(1,5)),['accuracy','precision','recall','f1_score'])\n",
    "plt.xlabel('preformance')\n",
    "plt.ylabel('percentage')\n",
    "fig = plt.gcf()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The cosine similarity method reaches a pretty high predicting accuracy around 78 percent. This indicates that considering the user's historical movies' genres is a very useful and strong way to recommend movies to them."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Classic method"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### each user's average rating on each genre"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "# We calculate each user's average rating for each genre of movie\n",
    "user_rating = defaultdict(list)\n",
    "genre_rating = defaultdict(list)\n",
    "last_user = 1\n",
    "for index, row in train.iterrows():\n",
    "    if row[0] != last_user:\n",
    "        user_rating[last_user] = genre_rating\n",
    "        genre_rating = defaultdict(list)\n",
    "        last_user = row[0]\n",
    "    genres = row[4].split('|')\n",
    "    for gen in genres:\n",
    "        genre_rating[gen].append(row[3])  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "for user,rating in user_rating.items():\n",
    "    for key,value in rating.items():\n",
    "        rating[key] = [sum(value)/len(value)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "# integrate each row to form a dataframe\n",
    "frames = []\n",
    "for user,rating in user_rating.items():\n",
    "    data = pd.DataFrame.from_dict(rating)\n",
    "    frames.append(data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_user = pd.concat(frames,sort=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "# set the index and name of the dataframe\n",
    "df_user.set_index(pd.Series(list(user_rating.keys())),inplace=True)\n",
    "df_user.index.name = 'userId'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Adventure</th>\n",
       "      <th>Children</th>\n",
       "      <th>Fantasy</th>\n",
       "      <th>Drama</th>\n",
       "      <th>Mystery</th>\n",
       "      <th>Sci-Fi</th>\n",
       "      <th>Thriller</th>\n",
       "      <th>Crime</th>\n",
       "      <th>Action</th>\n",
       "      <th>Comedy</th>\n",
       "      <th>Romance</th>\n",
       "      <th>War</th>\n",
       "      <th>Horror</th>\n",
       "      <th>Musical</th>\n",
       "      <th>Western</th>\n",
       "      <th>Animation</th>\n",
       "      <th>IMAX</th>\n",
       "      <th>Film-Noir</th>\n",
       "      <th>Documentary</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>userId</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>1</th>\n",
       "      <td>6.904455</td>\n",
       "      <td>6.599082</td>\n",
       "      <td>6.714009</td>\n",
       "      <td>7.756397</td>\n",
       "      <td>7.564128</td>\n",
       "      <td>7.117860</td>\n",
       "      <td>7.541193</td>\n",
       "      <td>7.891915</td>\n",
       "      <td>7.174452</td>\n",
       "      <td>7.107212</td>\n",
       "      <td>7.167975</td>\n",
       "      <td>7.627261</td>\n",
       "      <td>7.252730</td>\n",
       "      <td>7.413226</td>\n",
       "      <td>8.072794</td>\n",
       "      <td>6.883448</td>\n",
       "      <td>7.153846</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>7.583858</td>\n",
       "      <td>6.800000</td>\n",
       "      <td>7.405844</td>\n",
       "      <td>6.955556</td>\n",
       "      <td>6.645124</td>\n",
       "      <td>7.249075</td>\n",
       "      <td>6.721317</td>\n",
       "      <td>7.605442</td>\n",
       "      <td>7.511103</td>\n",
       "      <td>7.201896</td>\n",
       "      <td>7.140797</td>\n",
       "      <td>7.233164</td>\n",
       "      <td>5.746337</td>\n",
       "      <td>6.358333</td>\n",
       "      <td>7.438098</td>\n",
       "      <td>6.800000</td>\n",
       "      <td>6.800000</td>\n",
       "      <td>7.331361</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>6.950747</td>\n",
       "      <td>6.582382</td>\n",
       "      <td>7.217240</td>\n",
       "      <td>7.478415</td>\n",
       "      <td>7.198036</td>\n",
       "      <td>6.572171</td>\n",
       "      <td>7.057244</td>\n",
       "      <td>7.602656</td>\n",
       "      <td>6.775073</td>\n",
       "      <td>7.170722</td>\n",
       "      <td>7.461883</td>\n",
       "      <td>8.083648</td>\n",
       "      <td>6.567031</td>\n",
       "      <td>7.332899</td>\n",
       "      <td>7.488070</td>\n",
       "      <td>7.273578</td>\n",
       "      <td>NaN</td>\n",
       "      <td>7.082474</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>6.950745</td>\n",
       "      <td>7.039604</td>\n",
       "      <td>6.492986</td>\n",
       "      <td>6.828523</td>\n",
       "      <td>6.844158</td>\n",
       "      <td>6.571817</td>\n",
       "      <td>6.757774</td>\n",
       "      <td>6.464301</td>\n",
       "      <td>6.509078</td>\n",
       "      <td>6.391294</td>\n",
       "      <td>7.382287</td>\n",
       "      <td>8.158996</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>6.818533</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>7.081901</td>\n",
       "      <td>6.969567</td>\n",
       "      <td>6.786151</td>\n",
       "      <td>7.365239</td>\n",
       "      <td>6.899794</td>\n",
       "      <td>7.163044</td>\n",
       "      <td>7.216021</td>\n",
       "      <td>7.757410</td>\n",
       "      <td>7.086818</td>\n",
       "      <td>7.129127</td>\n",
       "      <td>6.915229</td>\n",
       "      <td>8.053777</td>\n",
       "      <td>8.268519</td>\n",
       "      <td>7.122426</td>\n",
       "      <td>7.083804</td>\n",
       "      <td>7.114327</td>\n",
       "      <td>7.546939</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        Adventure  Children   Fantasy     Drama   Mystery    Sci-Fi  Thriller  \\\n",
       "userId                                                                          \n",
       "1        6.904455  6.599082  6.714009  7.756397  7.564128  7.117860  7.541193   \n",
       "2        7.583858  6.800000  7.405844  6.955556  6.645124  7.249075  6.721317   \n",
       "3        6.950747  6.582382  7.217240  7.478415  7.198036  6.572171  7.057244   \n",
       "4        6.950745  7.039604  6.492986  6.828523  6.844158  6.571817  6.757774   \n",
       "5        7.081901  6.969567  6.786151  7.365239  6.899794  7.163044  7.216021   \n",
       "\n",
       "           Crime    Action    Comedy   Romance       War    Horror   Musical  \\\n",
       "userId                                                                         \n",
       "1       7.891915  7.174452  7.107212  7.167975  7.627261  7.252730  7.413226   \n",
       "2       7.605442  7.511103  7.201896  7.140797  7.233164  5.746337  6.358333   \n",
       "3       7.602656  6.775073  7.170722  7.461883  8.083648  6.567031  7.332899   \n",
       "4       6.464301  6.509078  6.391294  7.382287  8.158996       NaN       NaN   \n",
       "5       7.757410  7.086818  7.129127  6.915229  8.053777  8.268519  7.122426   \n",
       "\n",
       "         Western  Animation      IMAX  Film-Noir  Documentary  \n",
       "userId                                                         \n",
       "1       8.072794   6.883448  7.153846        NaN          NaN  \n",
       "2       7.438098   6.800000  6.800000   7.331361          NaN  \n",
       "3       7.488070   7.273578       NaN   7.082474          NaN  \n",
       "4       6.818533        NaN       NaN        NaN          NaN  \n",
       "5       7.083804   7.114327  7.546939        NaN          NaN  "
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_user.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### the number of films each user watches in each genre"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "# We calculate the number of each genre of movies each user watches\n",
    "user_count = defaultdict(list)\n",
    "genre_count = defaultdict(int)\n",
    "last_user = 1\n",
    "for index, row in train.iterrows():\n",
    "    if row[0] != last_user:\n",
    "        user_count[last_user] = genre_count\n",
    "        genre_count = defaultdict(int)\n",
    "        last_user = row[0]\n",
    "    genres = row[4].split('|')\n",
    "    for gen in genres:\n",
    "        genre_count[gen] += 1 "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "# integrate each row to from a dataframe\n",
    "for user,count in user_count.items():\n",
    "    for key,value in count.items():\n",
    "        count[key] = [value]\n",
    "        \n",
    "frames = []\n",
    "for user,count in user_count.items():\n",
    "    data = pd.DataFrame.from_dict(count)\n",
    "    frames.append(data)\n",
    "\n",
    "df_count = pd.concat(frames,sort=False)\n",
    "df_count.set_index(pd.Series(list(user_count.keys())),inplace=True)\n",
    "df_count.index.name = 'userId'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Adventure</th>\n",
       "      <th>Children</th>\n",
       "      <th>Fantasy</th>\n",
       "      <th>Drama</th>\n",
       "      <th>Mystery</th>\n",
       "      <th>Sci-Fi</th>\n",
       "      <th>Thriller</th>\n",
       "      <th>Crime</th>\n",
       "      <th>Action</th>\n",
       "      <th>Comedy</th>\n",
       "      <th>Romance</th>\n",
       "      <th>War</th>\n",
       "      <th>Horror</th>\n",
       "      <th>Musical</th>\n",
       "      <th>Western</th>\n",
       "      <th>Animation</th>\n",
       "      <th>IMAX</th>\n",
       "      <th>Film-Noir</th>\n",
       "      <th>Documentary</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>userId</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>1</th>\n",
       "      <td>71.0</td>\n",
       "      <td>18.0</td>\n",
       "      <td>67.0</td>\n",
       "      <td>40.0</td>\n",
       "      <td>18.0</td>\n",
       "      <td>38.0</td>\n",
       "      <td>41.0</td>\n",
       "      <td>20.0</td>\n",
       "      <td>63.0</td>\n",
       "      <td>39.0</td>\n",
       "      <td>11.0</td>\n",
       "      <td>8.0</td>\n",
       "      <td>42.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>9.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>14.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>19.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>19.0</td>\n",
       "      <td>18.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>18.0</td>\n",
       "      <td>9.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>16.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>49.0</td>\n",
       "      <td>10.0</td>\n",
       "      <td>20.0</td>\n",
       "      <td>56.0</td>\n",
       "      <td>11.0</td>\n",
       "      <td>92.0</td>\n",
       "      <td>50.0</td>\n",
       "      <td>21.0</td>\n",
       "      <td>61.0</td>\n",
       "      <td>49.0</td>\n",
       "      <td>15.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>32.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1.0</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>7.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>12.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>11.0</td>\n",
       "      <td>10.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>17.0</td>\n",
       "      <td>11.0</td>\n",
       "      <td>11.0</td>\n",
       "      <td>26.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>8.0</td>\n",
       "      <td>15.0</td>\n",
       "      <td>7.0</td>\n",
       "      <td>14.0</td>\n",
       "      <td>24.0</td>\n",
       "      <td>15.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>8.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        Adventure  Children  Fantasy  Drama  Mystery  Sci-Fi  Thriller  Crime  \\\n",
       "userId                                                                          \n",
       "1            71.0      18.0     67.0   40.0     18.0    38.0      41.0   20.0   \n",
       "2            14.0       1.0      1.0   19.0      4.0    19.0      18.0    1.0   \n",
       "3            49.0      10.0     20.0   56.0     11.0    92.0      50.0   21.0   \n",
       "4             5.0       1.0      1.0    7.0      3.0     4.0      12.0    6.0   \n",
       "5            17.0      11.0     11.0   26.0      2.0     8.0      15.0    7.0   \n",
       "\n",
       "        Action  Comedy  Romance  War  Horror  Musical  Western  Animation  \\\n",
       "userId                                                                      \n",
       "1         63.0    39.0     11.0  8.0    42.0      3.0      4.0        9.0   \n",
       "2         18.0     9.0      6.0  4.0    16.0      2.0      2.0        1.0   \n",
       "3         61.0    49.0     15.0  5.0    32.0      5.0      3.0        4.0   \n",
       "4         11.0    10.0      4.0  1.0     NaN      NaN      1.0        NaN   \n",
       "5         14.0    24.0     15.0  1.0     1.0      8.0      2.0        6.0   \n",
       "\n",
       "        IMAX  Film-Noir  Documentary  \n",
       "userId                                \n",
       "1        1.0        NaN          NaN  \n",
       "2        1.0        1.0          NaN  \n",
       "3        NaN        1.0          NaN  \n",
       "4        NaN        NaN          NaN  \n",
       "5        3.0        NaN          NaN  "
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_count.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### We want to check whether user's highest rating genre matches genres that they watch most"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "# count how many  user's highest rating genre matches genres that they watch most\n",
    "count_match = 0\n",
    "for index, row in df_user.iterrows():\n",
    "    if row.idxmax() == df_count.loc[index].idxmax():\n",
    "        count_match += 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.00501002004008016"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "count_match/len(df_user)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# show the percentage in a pie figure\n",
    "labels = 'watches most in their highest rating','watches most but does not rates highest'\n",
    "sizes = [count_match,len(df_user)-count_match]\n",
    "colors = ['gold', 'lightcoral']\n",
    "explode = (0.5, 0)\n",
    "\n",
    "# Plot\n",
    "plt.pie(sizes, explode=explode, labels=labels, colors=colors,\n",
    "autopct='%1.1f%%', shadow=True, startangle=140)\n",
    "fig = plt.gcf()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "From the above two dataframes we created and the analysis we made, we can find that if a user watches a large number of a certain kind of movie, it does not necessarily mean that they will rate that genre highest."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Make movie recommendation based on user's highest average rating"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "# we append those genres that user rates highest to their favorite genre vector\n",
    "# Here we choose a threshold 7 because after some experiments, we find out that 7 gives the highest accuracy\n",
    "user_prefer_score = defaultdict(list)\n",
    "for index,row in df_user.iterrows():\n",
    "    for i in range(len(row)):\n",
    "        if row[i] > 7:\n",
    "            user_prefer_score[index].append(df_user.columns[i])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "answer = list(test.apply(lambda row: 1 if row.rating >= 6.0 else -1, axis=1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [],
   "source": [
    "# we make recommendation based on user's highest rated genres\n",
    "pred = []\n",
    "for index,row in test.iterrows():\n",
    "    if row[0] not in df_user.index:\n",
    "        pred.append(-1)\n",
    "    else:\n",
    "        if set(user_prefer_score[row[0]]) & set(row[4].split('|')):\n",
    "            pred.append(1)\n",
    "        else:\n",
    "            pred.append(-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.7307384430658395\n"
     ]
    }
   ],
   "source": [
    "acc = sum(np.array(pred) == np.array(answer)) / len(answer)\n",
    "print(acc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "# calculate the precision,recall and f1_score of our model\n",
    "precision,recall,f_score,support = precision_recall_fscore_support(answer,pred,warn_for=())\n",
    "pre = precision[1]\n",
    "rec = recall[1]\n",
    "f1 = f_score[1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# show the accuracy, precision, recall and f1_score in a pie figure\n",
    "plt.boxplot([[acc],[pre],[rec],[f1]])\n",
    "plt.xticks(list(range(1,5)),['accuracy','precision','recall','f1_score'])\n",
    "plt.xlabel('preformance')\n",
    "plt.ylabel('percentage')\n",
    "fig = plt.gcf()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can observe that making movie recommendation based on user's highest average rating also reaches a pretty high accuracy around 73 precent. That is to say, we have supported our hypothesis that if users love a certain genre of movies, they tend to rate highest among these movies."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Make movie recommendation based on the genres that user watches most"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "# We append 5 genres each user watches most to their favorite genre vector\n",
    "user_prefer_genre = defaultdict(list)\n",
    "for index,row in df_count.iterrows():\n",
    "    user_prefer_genre[index] = [df_count.columns[i] for i in sorted(range(len(row)), key=lambda i: row[i])[-5:]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [],
   "source": [
    "# we make movie recommendation based on the 5 genres they watch most\n",
    "pred = []\n",
    "for index,row in test.iterrows():\n",
    "    if row[0] not in df_user.index:\n",
    "        pred.append(-1)\n",
    "    else:\n",
    "        if set(user_prefer_genre[row[0]]) & set(row[4].split('|')):\n",
    "            pred.append(1)\n",
    "        else:\n",
    "            pred.append(-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.5986591955173104\n"
     ]
    }
   ],
   "source": [
    "acc = sum(np.array(pred) == np.array(answer)) / len(answer)\n",
    "print(acc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [],
   "source": [
    "# calculate the precision,recall and f1_score of our model\n",
    "precision,recall,f_score,support = precision_recall_fscore_support(answer,pred,warn_for=())\n",
    "pre = precision[1]\n",
    "rec = recall[1]\n",
    "f1 = f_score[1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# show the accuracy, precision, recall and f1_score in a pie figure\n",
    "plt.boxplot([[acc],[pre],[rec],[f1]])\n",
    "plt.xticks(list(range(1,5)),['accuracy','precision','recall','f1_score'])\n",
    "plt.xlabel('preformance')\n",
    "plt.ylabel('percentage')\n",
    "fig = plt.gcf()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can find that making movie recommendation based on the genres that user watches most gives a relatively low accuracy. This is a very interesting finding. That is to say, users do not necessarily love those movies most that they watch most. If they watch a lot of a certain genre of movies, they not necessarily give high rates to them. That contradicts our hypothoesis that if users love a genre of movies, they would watch more of them. This phenomenon is understandable since it is kind of hard for users to always watch a movie among their favorites. Most of the time the movies they watch are not their favorites."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Ethics and Privacy\n",
    "\n",
    "When it comes to privacy and ethical considerations, there are many factors that we have to consider in order for our project to meet these expectations. \n",
    " \n",
    "First and foremost, we insured that we are using information that is valid and not leaving out any information that would change the outcome. We want to make sure that we are not using any false information so that our project provides a more precise reflection. \n",
    " \n",
    "Secondly, since the datasets we employed in our project come from Kaggle, a public platform of free datasets, we are sure that we have permission from the websites that we are taking information from. We want the readers of our project to see that we had help from rating websites and to give them the credit that they deserve. This project goes through the necessary procedures to be fair to rating websites as well as have their trust. \n",
    "Thirdly, we need to make sure that we are upholding the privacy of the user’s information from the website. In order to do this, we took out any personal information such as name, age, gender, etc. and labeled them using an id system. This insures that we are only using the necessary information for the experiment without breaking the privacy of the website and its users. \n",
    "Lastly, we want to make sure that we are not breaking any laws as well as not hurt the users or website in any way. Overall, the project will provide information that will be beneficial towards people that want to see ratings correlate with their favorite type of film. If we are able to get approval from others, others may try to build up from our project.\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Conclusion and Discussion"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "With our information, we are able to determine that the users rated movies they liked would match their passion and rated the movies they did not like would not match their passion. \n",
    "\n",
    "In this project we posed and tested three hypotheses.\n",
    "The first hypothesis is that a user will rate their favorite genre higher than the average for that genre (their favorite genre being the genre they give the highest ratings compared to other genres), and rate genres lower than average in their least favorite genre (the genre they give the lowest ratings to) that the films belonging to the genre that a user rates most highly will  a user will tend to rate films higher than the average of all users. This hypothesis is supported by our analysis: we found this to be true for 65% of users.\n",
    "The second hypothesis is if a user loves a certain genre of film, he or she tends to rate this genre of movies more. Our analysis does not support this hypothesis: this is true for only 47% of users goes against this hypothesis as only a small percentage of people tend to watch a movie that is a genre that they love.\n",
    "\n",
    "\n",
    "\n",
    "In our analysis, we decided to do cosine similarity which allows us to *insert info about this* as well as using a classic method that calculates each user's average rating for each genre of movie and number of genres of movies each user rated.\n",
    " \n",
    "\n",
    "One limitation to our analysis lies in our choice of sample population from the user database. Rather than using a random sample of the users, we used the first 2000 users in the database - presumably the earliest adopters of the source movie rating website, who are more possibly more zealous movie fans. This bias in sampling may therefore leads to inaccuracy in the predictive model. As a result, our conclusions can only be generalized to the entire user base provided that one takes as given the assumption that user film rating behavior is independent from registration date on the website. Thus, next time when generating the train dataset, we would possibly try the random sampling that random selects users to construct the train model in order to minimize such bias. Besides, another limitation is that due to the absence of average ratings of each film in the dataset, we have to calculate the average rating from our sample. Therefore, as our calculation may differ from the algorithm of IMDB website, the actual movie average ratings may hence different. Consequently, we may try to find a more comprehensive dataset next time.\n",
    "\n",
    "An impact that this project will have is that it will help people understand that people will tend to rate movies higher if it falls under their passion. Furthermore, our findings, though still incomplete, is also helpful in real practices, and one of its applications is the movie recommendation system of the movie websites. They could consider the extent of significance of the predictive factors we studied and hence recommend films according to the users’ rating histroy.\n",
    "\n",
    "In the future, we can improve our project by taking more factors into consideration to create the movie recommendation system. For example, in this project, we only considered the genres in the method of cosine similarity. In fact, we can also take in the director of the movie, the original country of the movie as well as the plot of the movie into the vectors we compare. We can form bag of words based on these information and do natural language processing through tfidf feature to find the most important parts that determines user's favor."
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}