reco_final_project.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 1. Business Objectives"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<b>Overall objective</b>\n",
    "\n",
    "- Our overall objective is to build a production-quality movie recommendation pipeline, in order to `provide personalizied TOP-10 list of movies specific to our users`. \n",
    "- These recommended movies will be displayed in the section of `TOP 10 MOVIES THAT YOU MAY LIKE` in the carousel on the APP or website. \n",
    "- The goal of this new pipeline is to increase the chance of users clicking into, watching and liking one of the recommended movies, thus improving user retention and activeness on our APP.\n",
    "\n",
    "\n",
    "<b>Target Users</b>\n",
    "\n",
    "- This pipeline will mainly target at `active users`, i.e., those with more than 20 ratings, to alleviate the cold start problem. \n",
    "- In the future, differnet pipelines, such as user survey, may be developed for movie recommendation for cold start users.\n",
    "\n",
    "\n",
    "<b>Sampling methods</b>\n",
    "\n",
    "- To improve recommendation accuracy, we use `Random Sampling after Minimum Requirement(RSMR)` to sample from the full MovieLens dataset, i.e., sample from the subset with cold users and movies excluded. \n",
    "- The final sample includes 1,170,000+ ratings applied to `1,000 movies` by `20,000 users`\n",
    "\n",
    "\n",
    "<b>Critical metrics that we look to optimize</b>\n",
    "- Most attention will be paid to improve the recommendation pipeline's performance on metrics like `recall@10` , `user coverage`: the higher these metrics are, the better job we are doing in making well recommendations to our active users. \n",
    "\n",
    "- We will also look at ranking accuracy metrics like `NDCG`, coverage metrics like `movie coverage`, `catalog coverage`, and how well our pipeline performs on different user groups of different rating levels (`recall over users`).\n",
    "\n",
    "- Tradeoffs we are willing to make: since our overal objective is to make good recommendation to users, we will care less on metrics like `movie-coverage`, `catalog-coverage` and `recall over movies`\n",
    "\n",
    "\n",
    "<b>Performance Requirements</b>\n",
    "\n",
    "- The system should be scalable with the selected dataset. Training time and testing time should be less than a certain level to allow for a daily-based update.\n",
    "\n",
    "\n",
    "<b>Recommendation Pipeline Architectures</b>\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![jupyter](./figure1.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 2. Data Exploration and Sampling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# basic packages\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import math\n",
    "import time\n",
    "from collections import defaultdict,OrderedDict\n",
    "from scipy import sparse\n",
    "np.warnings.filterwarnings('ignore')\n",
    "\n",
    "# evaluation\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn.metrics import make_scorer\n",
    "from sklearn.neighbors import KDTree\n",
    "\n",
    "# visualization\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "import seaborn as sns\n",
    "\n",
    "# self-build modules\n",
    "from model.bias_model import *\n",
    "from model.ann_kdtree import *\n",
    "from model.mf_nn import *\n",
    "from model.ann_nn import *\n",
    "from model.helper import *"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.1 Load & Understand the Full Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x576 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df = pd.read_csv('./ml-latest/ratings.csv')\n",
    "df_tag = pd.read_csv('./ml-latest/genome-scores.csv')\n",
    "\n",
    "#describe full rating dataset\n",
    "user_freq, movie_freq = describe(df)\n",
    "user_grouped = group(df, to_group = 'user')\n",
    "user_cnt = user_grouped.groupby(['groupId', 'groupRange']).count().rename(columns={\"userId\":\"count\"}).reset_index()\n",
    "movie_grouped = group(df, to_group = 'movie')\n",
    "movie_cnt = movie_grouped.groupby(['groupId', 'groupRange']).count().rename(columns={\"movieId\":\"count\"}).reset_index()\n",
    "plot_df(user_freq, movie_freq, user_cnt, movie_cnt, user_grouped, movie_grouped)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.2 Sampling & Train Test Spliting"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "`User and Movie threshold`: from above, a majority of users have rated more than 10-20 movies in our dataset, so we decide to focus on recommendating to these `active users (>20 ratings)` at this stage, to improve recommending recall and alleviate the cold start problem. Similar threshold was applied to movie as well to alleviate this problem. \n",
    "\n",
    "\n",
    "We then do a random sample to obtain a subset of the data, which contains `1,170,000+ ratings` applied to `1,000 movies` by `20,000 users`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of users: 20000\n",
      "number of items: 1000\n",
      "number of ratings: 1179969\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x576 with 4 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_sample = sample_df(df)\n",
    "\n",
    "#describe sample dataset\n",
    "user_freq_sample, movie_freq_sample = describe(df_sample)\n",
    "user_grouped_sample = group(df_sample, to_group = 'user')\n",
    "user_cnt_sample = user_grouped_sample.groupby(['groupId', 'groupRange']).count().rename(columns={\"userId\":\"count\"}).reset_index()\n",
    "movie_grouped_sample = group(df_sample, to_group = 'movie')\n",
    "movie_cnt_sample = movie_grouped_sample.groupby(['groupId', 'groupRange']).count().rename(columns={\"movieId\":\"count\"}).reset_index()\n",
    "plot_df(user_freq_sample, movie_freq_sample, user_cnt_sample, movie_cnt_sample, user_grouped_sample, movie_grouped_sample)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Train Test Split"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For each user, use the latest 20% ratings as testing set and oldest 80% as training set. to emulate the real-world movie-recommendating business setting."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "952005\n",
      "227964\n",
      "(1000, 20000)\n",
      "(1000, 20000)\n"
     ]
    }
   ],
   "source": [
    "train_df, test_df = train_test_split_by_time(df_sample)\n",
    "print(len(train_df))\n",
    "print(len(test_df))\n",
    "\n",
    "rating_train = train_df.pivot(index='movieId', columns='userId', values='rating')\n",
    "rating_test = test_df.pivot(index='movieId', columns='userId', values='rating')\n",
    "print(rating_train.shape)\n",
    "print(rating_test.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 3. Baseline Model for Benchmarks"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For part 3, two baseline models are built for benchmarking. One is `Bias Model`, another one is `Matrix Factorization Model(CF)`"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.1 Baseline Model 1: Bias Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train_time:  0.7887649536132812\n",
      "test_time:  0.38976287841796875\n"
     ]
    }
   ],
   "source": [
    "#model initializing\n",
    "model_bias = bias_model() \n",
    "\n",
    "#training\n",
    "start = time.time()    \n",
    "model_bias.fit(train_df)\n",
    "end = time.time()\n",
    "train_time = end - start\n",
    "\n",
    "#testing\n",
    "start = time.time()\n",
    "rating_pred_bias = model_bias.pred(test_df)\n",
    "end = time.time()\n",
    "test_time = end - start\n",
    "print(\"train_time: \", train_time)\n",
    "print(\"test_time: \", test_time)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean Recall: 0.032\n",
      "Recall for each user group:  {'10-50': 0.01, '50-100': 0.0126, '100-500': 0.0195, '500-1000': 0.0}\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "User coverage:  the fraction of users for which at least 1 items can be recommended well is 0.201\n",
      "Item coverage: the fraction of items that can be recommended to at least 1 users well is 0.033\n",
      "Catalog coverage: the fraction of items that are in the top-10 for at least 1 user is 0.033\n",
      "\n",
      "NDCG: 0.328\n"
     ]
    }
   ],
   "source": [
    "# Evaluation\n",
    "\n",
    "k_items = 10\n",
    "rating_pred = rating_pred_bias\n",
    "\n",
    "#recall\n",
    "user_recalls, avg_recall = recall_over_user(rating_test, rating_pred, k_items=k_items)\n",
    "print(f\"Mean Recall: {avg_recall:.3f}\")\n",
    "user_grouped = group(train_df, 'user')\n",
    "recall_dict = recall_over_user_group(rating_test, rating_pred, user_grouped=user_grouped, k_items=k_items)\n",
    "\n",
    "#coverage\n",
    "user_cov = cal_coverage(rating_test, rating_pred, k_items=k_items, threshold=1, coverage_type='user')\n",
    "movie_cov = cal_coverage(rating_test, rating_pred, k_items=k_items, threshold=1, coverage_type = 'item')\n",
    "catalog_cov = cal_coverage(rating_test, rating_pred, k_items=k_items, threshold=1, coverage_type = 'catalog')\n",
    "\n",
    "#ndcg\n",
    "ndcg_score = ndcg(rating_test, rating_pred)\n",
    "print()\n",
    "print(f\"NDCG: {ndcg_score:.3f}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<b>Result Interpretation</b>: \n",
    "\n",
    "The bias model basically use movie mean rating plus user mean rating as predicted ratings, and we will recommend top 10 movies that are not seen yet to users.\n",
    "\n",
    "Most users rated less than 50 movies in train set, and even less on test set. This means it will be inevitably difficult to predict what ther users will rate as Top 10 among the 900+ movies that are not rated yet by the users. So we accept a relatively low mean recall@10. \n",
    "\n",
    "The recall@10 rate of the bias baseline model is `3.2%`, and the fraction of users for which at least 1 movie can be recommended well is `20.1%`, with a NDCG score of `0.328`. This means we can recommend at least 1 movie well to about 1/5 of our active users. \n",
    "\n",
    "However, this bias model performs very bad (recall almost to zero) in the most active user groups (rated more 500 movies). This might be caused by that the mechanism of the bias model to take average of user ratings would average out user preferences, which make it difficult to predict well and recommend well to them. \n",
    "\n",
    "This is undesireble. Using other stronger algorithms might predict better with this user group."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.2 Baseline Model 2: MF Model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The next baseline Model we build is Model-based Collaborative Filtering Model, using Matrix Factorization powered by PySpark."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pyspark\n",
    "from pyspark.sql import SparkSession\n",
    "from pyspark.sql.functions import *\n",
    "from pyspark.sql.window import Window\n",
    "from pyspark.sql.types import *\n",
    "from pyspark.ml.evaluation import RegressionEvaluator\n",
    "from pyspark.ml.recommendation import ALS\n",
    "from pyspark.ml.tuning import CrossValidator, ParamGridBuilder "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "#initialize spark \n",
    "spark = SparkSession.builder.appName(\"PySpark ALS Model\").getOrCreate() \n",
    "train = spark.createDataFrame(train_df)  \n",
    "test = spark.createDataFrame(test_df)  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train_time:  32.672394037246704\n",
      "test_time:  27.65398406982422\n"
     ]
    }
   ],
   "source": [
    "#build model\n",
    "\n",
    "#We acheive the max average recall of 0.044 with cross validation, by setting max_iter=25, regParam = 0.1, rank=10.\n",
    "als = ALS(maxIter=25, regParam=0.1, rank=10, userCol=\"userId\", itemCol=\"movieId\", ratingCol=\"rating\",\n",
    "          coldStartStrategy=\"drop\")\n",
    "\n",
    "#training\n",
    "start = time.time()    \n",
    "model_mf = als.fit(train)\n",
    "end = time.time()\n",
    "train_time = end - start \n",
    "\n",
    "#testing \n",
    "start = time.time()\n",
    "predictions = model_mf.transform(test)\n",
    "rating_pred_mf = get_rating_pred(model_mf, predictions, rating_train)\n",
    "end = time.time()\n",
    "test_time = end - start\n",
    "\n",
    "print(\"train_time: \", train_time)\n",
    "print(\"test_time: \", test_time)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean Recall: 0.044\n",
      "Recall for each user group:  {'10-50': 0.0264, '50-100': 0.0292, '100-500': 0.0349, '500-1000': 0.1333}\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "User coverage:  the fraction of users for which at least 1 items can be recommended well is 0.259\n",
      "Item coverage: the fraction of items that can be recommended to at least 1 users well is 0.296\n",
      "Catalog coverage: the fraction of items that are in the top-10 for at least 1 user is 0.296\n",
      "\n",
      "NDCG: 0.349\n"
     ]
    }
   ],
   "source": [
    "# Evaluation\n",
    "\n",
    "k_items = 10\n",
    "rating_pred = rating_pred_mf\n",
    "\n",
    "#recall\n",
    "user_recalls, avg_recall = recall_over_user(rating_test, rating_pred, k_items=k_items)\n",
    "print(f\"Mean Recall: {avg_recall:.3f}\")\n",
    "user_grouped = group(train_df, 'user')\n",
    "recall_dict = recall_over_user_group(rating_test, rating_pred, user_grouped=user_grouped, k_items=k_items)\n",
    "\n",
    "#coverage\n",
    "user_cov = cal_coverage(rating_test, rating_pred, k_items=k_items, threshold=1, coverage_type='user')\n",
    "movie_cov = cal_coverage(rating_test, rating_pred, k_items=k_items, threshold=1, coverage_type = 'item')\n",
    "catalog_cov = cal_coverage(rating_test, rating_pred, k_items=k_items, threshold=1, coverage_type = 'catalog')\n",
    "\n",
    "#ndcg\n",
    "ndcg_score = ndcg(rating_test, rating_pred)\n",
    "print()\n",
    "print(f\"NDCG: {ndcg_score:.3f}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<b>Result Interpretation</b>: \n",
    "\n",
    "The MF baseline model performs better in all metrics: `recall@10 = 4.4%`, `user coverage = 25.9%`, `NDCG = 0.349`, `movie coverage = 29.6%`.\n",
    "\n",
    "The mean recall also increases as the ratings of users increase. Especially, it has a relatively high recall on the most active users (`recall = 0.12`). This indicates that MF model tend to work well for the active users with a great number of ratings, the density of which help the model figure out the latent features well."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 4. ANN Model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For part 4, we will explore another efficient algorithm, `Approximate Nearest Neighbors (ANN)`, using user and movie embeddings genereated by the MF model. Since both users and movies are embedded in the same dimension space, we only need to find the nearest 10 movies from the user that are not rated by the user yet. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "training begins ......\n",
      "Begin to retrieve latent features from cf ......\n",
      "training finishes.\n",
      "prediction begins ......\n",
      "prediciton finishes.\n",
      "train_time:  49.69563293457031\n",
      "test_time:  24.325652837753296\n"
     ]
    }
   ],
   "source": [
    "#model initializing\n",
    "\n",
    "#We acheive the max average recall of 0.053 with cross validation, by setting max_iter=5, rank=20, leaf_size=50.\n",
    "k_items = 10\n",
    "max_iter = 5\n",
    "rank = 20\n",
    "leaf_size = 50\n",
    "\n",
    "#training\n",
    "start = time.time()  \n",
    "model_ann = ann_kdtree(max_iter=max_iter, regparam=0.1, rank=rank, leaf_size=leaf_size, k_items=k_items)\n",
    "model_ann.fit(train_df)\n",
    "end = time.time()  \n",
    "train_time = end - start\n",
    "\n",
    "#testing\n",
    "start = time.time()  \n",
    "rating_pred_ann =  model_ann.pred(test_df)\n",
    "end = time.time() \n",
    "test_time = end - start \n",
    "\n",
    "print(\"train_time: \", train_time)\n",
    "print(\"test_time: \", test_time)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean Recall: 0.053\n",
      "Recall for each user group:  {'10-50': 0.032, '50-100': 0.0355, '100-500': 0.0357, '500-1000': 0.0667}\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "User coverage:  the fraction of users for which at least 1 items can be recommended well is 0.307\n",
      "Item coverage: the fraction of items that can be recommended to at least 1 users well is 0.444\n",
      "Catalog coverage: the fraction of items that are in the top-10 for at least 1 user is 0.444\n",
      "\n",
      "NDCG: 0.324\n"
     ]
    }
   ],
   "source": [
    "#evaluation\n",
    "\n",
    "k_items = 10\n",
    "rating_pred = rating_pred_ann\n",
    "\n",
    "#recall\n",
    "user_recalls, avg_recall = recall_over_user(rating_test, rating_pred, k_items=k_items)\n",
    "print(f\"Mean Recall: {avg_recall:.3f}\")\n",
    "user_grouped = group(train_df, 'user')\n",
    "recall_dict = recall_over_user_group(rating_test, rating_pred, user_grouped=user_grouped, k_items=k_items)\n",
    "\n",
    "#coverage\n",
    "user_cov = cal_coverage(rating_test, rating_pred, k_items=k_items, threshold=1, coverage_type='user')\n",
    "movie_cov = cal_coverage(rating_test, rating_pred, k_items=k_items, threshold=1, coverage_type = 'item')\n",
    "catalog_cov = cal_coverage(rating_test, rating_pred, k_items=k_items, threshold=1, coverage_type = 'catalog')\n",
    "\n",
    "#ndcg\n",
    "ndcg_score = ndcg(rating_test, rating_pred)\n",
    "print()\n",
    "print(f\"NDCG: {ndcg_score:.3f}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<b>Result Interpretation</b>: \n",
    "\n",
    "The ANN model further improves performances in most metrics: `recall@10 = 5.3%`, `user coverage = 30.7%`, `NDCG = 0.324`, `movie coverage = 44.4%`.\n",
    "\n",
    "ANN model also predicts better as ratings of movie increase for each user group. Although `recall(0.07)` for the most active user group is lower than `MF model(0.13)`, it performs  better for the majority group (group 10-50, 50-100, 100-500)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 5. Hybrid Model Exploration "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For part 5, we will try combinations of different algorithms, such as MF, ANN, NN etc, to develop a stronger hybrid pipeline model.\n",
    "\n",
    "The hope is that introducing more complicated structures like NN in reranking stage will help improve recommendation metrics(recall@10)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 5.1 MF-NN Model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Algorithms deployed:\n",
    "\n",
    "- Preranking stage: MF \n",
    "- Reranking stage: NN"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 5.1.1 MF-NN Model (MF embedding)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training begins.......\n",
      "training set created\n",
      "start training neural network......\n",
      "model training finished\n",
      "creating testing set for neural network.....\n",
      "prediction finished\n",
      "train_time:  398.08773398399353\n",
      "test_time:  50.552711963653564\n"
     ]
    }
   ],
   "source": [
    "#model initializing\n",
    "tuned_model_mfnn = mf_nn(max_iter = 15, regparam = 0.1, k = 30, rank = 10, movie_tag_embed=False) \n",
    "\n",
    "#training\n",
    "start = time.time()    \n",
    "tuned_model_mfnn.fit(train_df, df_tag)\n",
    "end = time.time()\n",
    "train_time = end - start\n",
    "\n",
    "#testing\n",
    "start = time.time()\n",
    "rating_pred_mfnn = tuned_model_mfnn.pred(test_df)\n",
    "\n",
    "end = time.time()\n",
    "test_time = end - start\n",
    "print(\"train_time: \", train_time)\n",
    "print(\"test_time: \", test_time)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_history(tuned_model_mfnn.history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean Recall: 0.029\n",
      "Recall for each user group:  {'10-50': 0.0178, '50-100': 0.0203, '100-500': 0.0205, '500-1000': 0.0667}\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXwAAAEaCAYAAAASSuyNAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8QZhcZAAAgAElEQVR4nO3df1zP9/7/8VuSCC2st7Yym7GcjYRMfnwybOVHb/3w45goQ/vhmMlZhvycNrSOTLFNHIyM1igN6czMOZaxmtGOZmImTT8IRaVfr+8fTq+vt8q7Jr/2elwvF5dLr9fz9Xq+n8/n++X+fvV8v9/PTBRFURBCCPGn1+B+N0AIIcS9IYEvhBAaIYEvhBAaIYEvhBAaIYEvhBAaIYEvhBAaIYF/HwwYMAB7e3uDf126dGHw4MGsX7/+nrbF3t6euLg4AGbOnMn48eNve/zVq1f58MMPGTRoEJ07d6ZPnz5MmTKFI0eO3IPWGtq2bZvBGP7lL3/ByckJX19fvv32W4Njx40bR1BQkLq9evVqevbsSdeuXfnpp5/4z3/+w4ABA+jcuTOffvrpve6K6vLly8TExNRYvm3bNp599tk6l91t//nPf3j99dfp06cPnTp1YsCAASxYsIDs7Oz70h5RvYb3uwFa5e/vj5+fn7p9+fJltmzZwuLFi9HpdAwZMuQ+tq56ly9fZuzYsZSWlvLWW2/h4OBAXl4e0dHRjB07lnfffZfhw4ff0zaZmpqyf/9+AMrLy7lw4QKfffYZkyZNIiIigoEDBwIQHh5Ow4Y3Lvdr166xbNkyXn/9dUaOHIlOp2P06NE89dRTfPrpp1hZWd3TPtwsNDSU3377jREjRty3NtTVqlWrWLlyJePHj+fNN9+kRYsWnDlzho8//pjhw4cTGxvLo48+er+bKZDAv28sLCywtrZWt62trZk7dy7//ve/2bVr1wMZ+IsWLeLatWts375dDUU7OzscHBx49NFHWbhwId27d+fJJ5+8p+26eRxtbGx47733yMvLY9GiRbi4uGBmZmYQ4levXkVRFJydnbG1tQWgoKCAfv36YWdnd0/bfquH7XuQR44cYcWKFSxcuJC//vWv6v7HH3+cHj164O7uztq1a3nnnXfuYytFJZnSecCYmZmpd6IA58+fZ+rUqXTr1o3evXsTEBBg8GuyoiisX78eV1dXunTpgoeHh3rHC7Bnzx6GDx+Og4MDXbp0YfTo0Rw7dqzO7crLy2P37t34+flVewc8efJkzMzMiI6OJiMjg44dO3Lo0CGDY/z9/Xn77bcBuHLlCrNmzaJnz548//zz+Pv7c/r0afXYmTNnMm3aNMaNG0f37t3ZvHlzndrr6+vL+fPn+eGHH4D/P6Vz6NAhXFxcAPDz82PcuHHY29vz22+/sXLlSuzt7QEoKSlhyZIl9O3bl27dujF27Fh+/PFHtf7w8HDGjRunPjdhYWEAfPXVVwwbNozOnTszaNAg1q5dS0VFBQDnzp3D3t6ePXv24OXlpT5fX331lVpnTEwMhw8fxt7ennPnztWpz7f68ccfGT16NI6OjvTs2ZPAwEAuX76slv/yyy9MnDiRLl264OLiwrx588jPz1fLBwwYwNKlS3Fzc8PZ2Zn//ve/VR5j06ZNtG3bllGjRlUpMzMzY926dQQEBBj0/+OPP6ZXr14MHjyYkpISfv/9dwICAujVqxddu3Zl8uTJZGRkGLRj1apVBnXfvC88PJzx48ezfPlyevToQc+ePQkODqakpEQ9fvXq1QwcOJBOnTrh5uZGVFTUHxzVh5sE/gOiqKiINWvWcOrUKfR6PQCFhYWMGzcOc3NztmzZwtq1ayktLcXPz0+9mCMjI1mxYgWTJ08mPj6eQYMG8be//Y2TJ09y7Ngxpk2bhre3N7t27WLjxo0AzJ07t87tS01Npby8nG7dulVb3qhRIxwdHTly5Aht2rShe/fu7Ny5Uy3Py8sjKSkJLy8vFEXh1VdfJScnhzVr1rB582Yef/xxxowZw6VLl9Rzdu/ezUsvvUR0dDQvvfRSndpbGdy//PKLwf6uXbuyfft24EZQhIeHc+DAAWxtbZkwYQIHDhwAYMaMGXz//fcsX76cL774AmdnZ3x9ffn111/Vug4fPkybNm3Yvn07I0aMYP/+/bz99tv4+vqyc+dOAgMD+fTTT6uEVUhICAEBAXz++ec89thjvPPOOxQWFjJhwgTc3d3p2rUrBw4c4LHHHqtTn29WXl7OG2+8Qa9evfjyyy9ZvXo1qampLF26FIDs7GzGjRvHM888w/bt21mxYgXp6elMmTLFoJ7PPvuMRYsW8cknn/CXv/ylyuOkpKTQs2dPTExMqm3H448/TqNGjQz27dy5k02bNhEaGkpJSQkvv/wyV65cYc2aNWzcuJGCggLGjh1LQUFBrfubnJzMDz/8wMaNGwkJCSEhIYHg4GAAvv76a9auXUtwcDB79uxh0qRJLFq0iO+//77W9f9ZyJTOfbJq1SoiIyOBG3fp169fx97enmXLlqnzzjt37qSoqIglS5ZgamoKwLJly+jZsyeJiYkMHTqUTz/9lFdeeQVPT08A3njjDcrKyigsLKRRo0bMnz+f0aNHAzemX0aOHMmcOXPq3N7KO7/bzW9bWVmpd6Wenp6EhoYyd+5czMzM2L17N61ataJXr14cPHiQ1NRUDh8+TLNmzQBYuHAh3333HdHR0bz22mvAjakaX1/fOrcVwNLSErgxfXOzRo0a0bJlSwAeeeQRtT+mpqbqNNtvv/3G7t27+fLLL+nQoQMAU6ZMISUlhXXr1vHuu+8CYGJiwptvvknjxo2BGy8SL7/8sjr//sQTT3Dt2jXmzp3L5MmT1TZMnDhR/S3jrbfewtPTk/T0dBwcHGjcuDFmZmYG01R/REFBAZcuXeLRRx/F1tYWOzs7Vq5cSWlpKQCbN2/Gzs7OYKolLCwMFxcXjhw5QteuXYEbd9LPP/98jY9z4cIFWrRoYbDv3XffVV9U4Ubo3/zi7+Pjw9NPPw1AVFQU+fn5LFu2TH0uPvzwQwYMGMCOHTvw8fGpVX9NTU0JCwujVatWdOzYkWnTprFgwQJmzJjB2bNnMTMz4/HHH8fW1paRI0diZ2dHu3btalX3n4kE/n3i4+PDmDFjKC8vZ+/evaxatQpvb2+GDh2qHnP8+HHy8vJwcnIyOLeoqIhTp05x6dIlcnNzcXBwMCh/88031Z+bN2/OJ598Qnp6Or/99htpaWnqFENdVP5nvDVAb1ZQUKCG6eDBgwkODubbb7/lhRdeID4+nmHDhtGgQQOOHz9OeXk5//d//2dw/vXr1zl16pS6fSfz6ZXtrAz+ujh+/DhAlWmKkpISg2kCa2trNewB0tLSSE1NZcuWLeq+iooKiouLyczMVO+Cn3rqKbW8efPmAGoQG9OwYcMa5/krKirU6UArKyteeeUV3n33XcLDw+nTpw/9+/dn8ODBalvT0tLUYL/ZqVOn1P1t2rS5bXusrKwMpongxotj5QcSNm/ezL/+9S+D8pvrPHnyJO3atTO4kWjZsiVPP/10ld/Obqddu3a0atVK3XZ0dKS0tJRff/0VvV5PTEwMrq6uPPPMM/Tt25dhw4YZHK8VEvj3ySOPPELbtm2BGxdrgwYNeO+992jZsiXu7u7AjTnQ9u3bExERUeX85s2bY2ZmdtvH+O677/D392fgwIF069aN4cOHc+bMGebPn1/n9jo4OGBmZkZKSkq1H/0rKSnh6NGjeHt7A9CsWTNefPFFvvzyS55++mmOHDmi/opd+SZqdHR0lXosLCzUn28O07qqDO0/8jHFynHdsmVLlTbcPD1xa5mZmRmTJk1Sp+Ru1rp1a3Jycgzqv1lt36x95JFHqKio4Nq1azRt2tSgLD8/n0ceeUTdfuedd/Dx8WH//v0cOHCAWbNmsWPHDlavXo2ZmRl9+vSp9re9yhdtAHNz89u2p1u3bqSkpFQ5/+bfom51c5011V9RUXHb67usrMxg++b3veDGlBZAgwYNaNWqFTt27CAlJYUDBw6wf/9+NmzYwNKlS6t9rv7MZA7/ATFhwgS6d+/OwoULyc3NBaBDhw6cO3cOKysr2rZtS9u2bWnVqhWLFy/ml19+oXnz5lhbW5OammpQ17hx49S58T59+rB8+XJ8fX1xdnYmMzMTqPunQR555BG8vLxYu3atwTx7pbVr11JYWGhwV+zl5cW+ffuIj4+nU6dOtG/fXu1X5V1hZb/s7OxYvnx5vc2rbt68mTZt2tClS5c6n1s5jXPx4kW1fW3btmX9+vXs3bu3xvPat2/PmTNnDM755Zdf1Dd0a6OmufBKlS9gt4Zs5b7OnTsDcPbsWebPn4+1tTU+Pj589NFHLF26lP3793Px4kXat2/PqVOnePzxx9W2NmjQgPfff5/z58/Xur3jxo0jPT2d2NjYasuzsrJue3779u05ffq0wW8JeXl5/Prrr+q0j5mZmcFvllevXuXixYsG9fz6668UFhaq20ePHqVx48a0a9eOXbt28dlnn9GjRw8CAgKIjY2lT58+7Nixo9b9/LOQwH9AmJiYsGjRIoqLi9U7Yb1eT4sWLZg2bRqpqan88ssv/P3vf+fo0aNqKE2aNIn169ezc+dOzp49y6pVqzh69Cj9+vWjZcuWnDhxgh9//JGMjAw2btzIhg0bAAymJmrrnXfeUT+znpCQQGZmJmlpaQQHBxMREcH8+fMNpit69+5N06ZNWbNmjfoeA0CvXr1wdHRk2rRpJCcn8+uvvzJnzhz27dvHM888U+d25ebmkpubS3Z2NqmpqcyYMYOvv/6a+fPn06BB3S/xtm3bMmTIEObOncv+/fs5e/YsYWFhbNmyRQ2h6rzxxhvs3LmT1atXc+bMGb755hvmzZtH48aNq7xxWZOmTZuSnZ1NRkZGlbtYuDGN5OHhwdy5c0lMTCQzM5OffvqJkJAQ/v3vf+Pv7w9AixYt2L17NwsWLODUqVOcOnWK3bt388QTT9CiRQvGjh1Lfn4+M2fO5MSJE6SmpjJ9+nTOnDlTp4/V9ujRg+nTpxMUFERwcDBHjx4lMzOTAwcO8MYbb7B161Z69uxZ4/nDhg2jZcuWTJ8+nePHj/Pf//6X6dOnY2lpqU5vOjo6snPnTo4cOcLJkyeZOXOm+p5WpYKCAmbNmkV6ejr79u1j+fLljBkzhiZNmlBSUsLSpUvZsWMHmZmZHDx4kOPHj/+hm4GHnUzpPECefvppXnvtNcLDw9m7dy8DBw5k3bp1LFmyBD8/P0xMTHB0dGTDhg3q/KOvry/FxcV88MEH5OXl0aFDBz7++GM6dOjA1KlTycnJYeLEiZiammJvb8+SJUsICAggNTW1ynsDxjRr1oyNGzfy6aefsnLlSs6ePUvTpk1xcnIiKioKR0dHg+MbNGjAsGHDWL9+vcF7EyYmJqxcuZKlS5cyefJkSkpK+Mtf/sKaNWvU3wJqq7y8nL59+wI33rhr1aoVDg4ObNq0qcZPFNVGcHAw//jHP5g9ezYFBQU8/fTThIeH06tXrxrPcXFxISQkhNWrV7NixQpatmyJp6en+rHE2vD29uarr75iyJAhREVFVXl/BuC9997jk08+YdmyZfz+++80btyYTp06sW7dOnXuvXnz5kRGRvLBBx8watQoKioq6NGjB6tXr6ZBgwZYW1uzbt06QkNDGTVqFI0bN6Znz558+OGHtX5xqvTqq6/SrVs3Nm3axJQpU7h06RJWVlY4OTmxYcMGnJ2dazzX3NyctWvXsmTJEnx8fDA1NaVXr15ERUWp779Mnz6defPmMX78eJo3b86ECROqvJdkZ2fHE088wahRo7CwsOCvf/2r+okjT09PLl68SHh4OOfPn6dVq1Z4e3vz+uuv16mffwYm8hevhBAPs/DwcHbs2FHlzWFRlUzpCCGERtRqSic+Pp6PPvqIsrIy/Pz8qnw2Ni0tjaCgIK5du4aTkxMLFy7kypUrTJgwQT2m8nPB92ORLSGEELWY0snOzubll19m27ZtNGrUiNGjR7Ns2TKDuVZ3d3eCg4NxdHRk9uzZdOrUiTFjxqjlFRUV+Pn5MWrUKM19DEoIIR4URqd0kpKScHZ2xsrKCgsLC9zc3EhISFDLMzMzKS4uVt+w8/b2NigH+OKLL2jSpImEvRBC3EdGp3RycnIMvuat0+kMFt+6tdza2tpgca/y8nI+/vjjKuuJVMrPzzdYsAlufGQwIyODJ598ssrHr4QQQlSvvLyc3NxcOnXqVO0XF40GfkVFhcGXQRRFMdg2Vv6f//yHJ598Ul3M6lYbNmyo9pukQggh/pioqKhqP3ZtNPBtbGxITk5Wt3Nzc9HpdAblld8MhRuLKd1cXvmZ4pr4+fnh5eVlsC8zMxNfX1+ioqKwsbEx1kQhhBDc+Gazj49PjYvvGQ383r17Ex4eTl5eHk2aNCExMZFFixap5ba2tpibm5OSkkL37t2Ji4tTVwKEG2tyV377rzqWlpY1LnBlY2Nz3/8ghRBCPGxqmgo3+qZt69atCQgIwNfXF09PT9zd3XFwcMDf319dwyU0NJTFixczaNAgCgsLDZa0zcjIkLt0IYR4ADyQ37Q9d+4cAwcOZO/evXKHL4QQtWQsO+WbtkIIoRES+EIIoRES+EIIoRES+EIIoRES+EIIoRES+EKIB155Se3+yLsW3MlYyF+8EkI88EwbmbHL95X73YwHwpBP1/3hc+UOXwghNEICXwghNEICXwghNEICXwghNEICXwghNEICXwghNEICXwghNEICXwghNEICXwghNEICXwghNEICXwghNEICXwghNEICXwghNEICXwghNEICXwghNKJWgR8fH8+QIUNwdXUlKiqqSnlaWhre3t64ubkRFBREWVkZADk5Obz66qt4enoyevRozp07V7+tF0IIUWtGAz87O5uwsDA2b95MbGwsW7duJT093eCYwMBA5s2bx549e1AUhejoaABmzJhB//79iY2NxcPDg9DQ0LvTCyGEEEYZDfykpCScnZ2xsrLCwsICNzc3EhIS1PLMzEyKi4txdHQEwNvbm4SEBPLy8vj5558ZPXo0AMOHD2fatGl3qRtCCCGMMfonDnNycrC2tla3dTodx44dq7Hc2tqa7OxsMjIyePzxx1myZAnJyclYW1szd+7cKvXn5+eTn59vsC8rK+sPdUYIIUTNjAZ+RUUFJiYm6raiKAbbNZWXlZVx/Phx3nzzTWbNmsXnn3/OzJkz2bhxo0H9GzZsICIioj76IoQQ4jaMBr6NjQ3Jycnqdm5uLjqdzqA8NzdX3b5w4QI6nQ5ra2uaNm1K//79AXB3dyc4OLhK/X5+fnh5eRnsy8rKwsfHp+69EUIIUSOjc/i9e/fm4MGD5OXlUVRURGJiIi4uLmq5ra0t5ubmpKSkABAXF4eLiwtPPPEENjY27N+/H4B9+/bx3HPPVanf0tISOzs7g382Njb11T8hhBD/YzTwW7duTUBAAL6+vnh6euLu7o6DgwP+/v6kpqYCEBoayuLFixk0aBCFhYX4+voCEB4ezpo1a3B3d+fTTz/l/fffv7u9EUIIUSOjUzoAer0evV5vsC8yMlL9uWPHjsTExFQ5r127dlXm7IUQQtwf8k1bIYTQCAl8IYTQCAl8IYTQCAl8IYTQCAl8IYTQCAl8IYTQCAl8IYTQCAl8IYTQCAl8IYTQCAl8IYTQCAl8IYTQCAl8IYTQCAl8IYTQCAl8IYTQCAl8IYTQCAl8IYTQCAl8IYTQCAl8IYTQCAl8IYTQCAl8IYTQCAl8IYTQCAl8IYTQiFoFfnx8PEOGDMHV1ZWoqKgq5WlpaXh7e+Pm5kZQUBBlZWUAbN++nb59++Lh4YGHhwdhYWH123ohhBC11tDYAdnZ2YSFhbFt2zYaNWrE6NGj6dmzJ+3bt1ePCQwMJDg4GEdHR2bPnk10dDRjxozhp59+YubMmbi7u9/VTgghhDDO6B1+UlISzs7OWFlZYWFhgZubGwkJCWp5ZmYmxcXFODo6AuDt7a2Wp6amsn37dvR6PW+//TZXrly5S90QQghhjNHAz8nJwdraWt3W6XRkZ2fXWG5tba2WW1tbM3nyZHbs2MFjjz3Gu+++W6X+/Px8zp07Z/AvKyvrjjolhBCiKqNTOhUVFZiYmKjbiqIYbN+ufOXKler+SZMm8dJLL1Wpf8OGDURERPyx1gshhKg1o4FvY2NDcnKyup2bm4tOpzMoz83NVbcvXLiATqejoKCAL774gvHjxwM3XghMTU2r1O/n54eXl5fBvqysLHx8fOrcGSGEEDUzOqXTu3dvDh48SF5eHkVFRSQmJuLi4qKW29raYm5uTkpKCgBxcXG4uLhgYWHBmjVrOHr0KACbNm2q9g7f0tISOzs7g382Njb11T8hhBD/Y/QOv3Xr1gQEBODr60tpaSkjRozAwcEBf39/pk6dSufOnQkNDWXOnDlcvXqV5557Dl9fX0xNTVm+fDkLFiyguLiYJ598kpCQkHvRJyGEENUwGvgAer0evV5vsC8yMlL9uWPHjsTExFQ5z8nJie3bt99hE4UQQtQH+aatEEJohAS+EEJohAS+EEJohAS+EEJohAS+EEJohAS+EEJohAS+EEJohAS+EEJohAS+EEJohAS+EEJohAS+EEJohAS+EEJohAS+EEJohAS+EEJohAS+EEJohAS+EEJohAS+EEJohAS+EEJohAS+EEJohAS+EEJohAS+EEJohAS+EEJoRK0CPz4+niFDhuDq6kpUVFSV8rS0NLy9vXFzcyMoKIiysjKD8uPHj9OpU6f6abEQQog/xGjgZ2dnExYWxubNm4mNjWXr1q2kp6cbHBMYGMi8efPYs2cPiqIQHR2tlhUVFbFo0SJKS0vrv/VCCCFqzWjgJyUl4ezsjJWVFRYWFri5uZGQkKCWZ2ZmUlxcjKOjIwDe3t4G5UuWLMHPz+8uNF0IIURdNDR2QE5ODtbW1uq2Tqfj2LFjNZZbW1uTnZ0NwN69eykuLmbQoEE11p+fn09+fr7BvqysrNr3QAghRK0YDfyKigpMTEzUbUVRDLZrKs/NzeWjjz5i/fr1t61/w4YNRERE/IGmCyGEqAujgW9jY0NycrK6nZubi06nMyjPzc1Vty9cuIBOp+Obb77h8uXL+Pj4qGUeHh5ERUXRrFkzdZ+fnx9eXl4Gj5mVlWVwnhBCiDtnNPB79+5NeHg4eXl5NGnShMTERBYtWqSW29raYm5uTkpKCt27dycuLg4XFxdGjhzJyJEj1ePs7e2Ji4urUr+lpSWWlpb11B0hhBA1MfqmbevWrQkICMDX1xdPT0/c3d1xcHDA39+f1NRUAEJDQ1m8eDGDBg2isLAQX1/fu95wIYQQdWP0Dh9Ar9ej1+sN9kVGRqo/d+zYkZiYmNvWceLEiT/QPCGEEPVFvmkrhBAaIYEvhBAaIYEvhBAaIYEvhBAaIYEvhBAaIYEvhBAaIYEvhBAaIYEvhBAaIYEvhBAaIYEvhBAaIYEvhBAaIYEvhBAaIYEvhBAaIYEvhBAaIYEvhBAaIYEvhBAaIYEvhBAaIYEvhBAaIYEvhBAaIYEvhBAaIYEvhBAaIYEvhBAaUavAj4+PZ8iQIbi6uhIVFVWlPC0tDW9vb9zc3AgKCqKsrAyA5ORkvL290ev1vP7661y5cqV+Wy+EEKLWjAZ+dnY2YWFhbN68mdjYWLZu3Up6errBMYGBgcybN489e/agKArR0dEAzJo1i5CQEOLj42nfvj1r1669O70QQghhlNHAT0pKwtnZGSsrKywsLHBzcyMhIUEtz8zMpLi4GEdHRwC8vb3V8l27dtG+fXtKS0vJzs7G0tLyLnVDCCGEMQ2NHZCTk4O1tbW6rdPpOHbsWI3l1tbWZGdnA2BmZsaJEyd45ZVXaNiwIdOnT69Sf35+Pvn5+Qb7srKy6t4TIYQQt2U08CsqKjAxMVG3FUUx2DZWbm9vT1JSElu2bCEgIIAtW7YY1L9hwwYiIiLuqBNCCCGMMzqlY2NjQ25urrqdm5uLTqersfzChQvodDquX7/OV199pe4fNmwYJ06cqFK/n58fe/fuNfhX3RvDQggh7ozRwO/duzcHDx4kLy+PoqIiEhMTcXFxUcttbW0xNzcnJSUFgLi4OFxcXGjYsCELFy7kp59+AmD37t1069atSv2WlpbY2dkZ/LOxsamv/gkhhPgfo1M6rVu3JiAgAF9fX0pLSxkxYgQODg74+/szdepUOnfuTGhoKHPmzOHq1as899xz+Pr6YmpqSlhYGPPmzaO8vJzWrVvz3nvv3Ys+CSGEqIbRwAfQ6/Xo9XqDfZGRkerPHTt2JCYmpsp5Tk5ObNu27Q6bKIQQoj7IN22FEEIjJPCFEEIjJPCFEEIjJPCFEEIjJPCFEEIjJPCFEEIjJPCFEEIjJPCFEEIjJPCFEEIjJPCFEEIjJPCFEEIjJPCFEEIjJPCFEEIjJPCFEEIjJPCFEEIjJPCFEEIjJPCFEEIjJPCFEEIjJPCFEEIjJPCFEEIjJPCFEEIjJPCFEEIjJPCFEEIjahX48fHxDBkyBFdXV6KioqqUp6Wl4e3tjZubG0FBQZSVlQGQkpLCiBEj8PDwwM/Pj8zMzPptvRBCiFozGvjZ2dmEhYWxefNmYmNj2bp1K+np6QbHBAYGMm/ePPbs2YOiKERHR6v7g4ODiYuLQ6/XExwcfHd6IYQQwiijgZ+UlISzszNWVlZYWFjg5uZGQkKCWp6ZmUlxcTGOjo4AeHt7k5CQQElJCW+99RYdO3YEwN7envPnz1epPz8/n3Pnzhn8y8rKqq/+iVqqKCu93014YNTHWJTIeKpkLB4cDY0dkJOTg7W1tbqt0+k4duxYjeXW1tZkZ2fTqFEjPDw8AKioqCAiIoIXX3yxSv0bNmwgIiLijjoh7lyDhmakhEy63814IHSfseaO62jU0Izx696qh9Y8/Na/8uH9boL4H6OBX1FRgYmJibqtKIrBtrHykpISZs6cSVlZGa+99lqV+v38/PDy8jLYl5WVhY+PT916IoQQ4raMBr6NjQ3Jycnqdm5uLjqdzqA8NzdX3b5w4YJafu3aNd544w2srKz46KOPMDMzq1K/paUllpaWd5CRkecAABQCSURBVNQJIYQQxhmdw+/duzcHDx4kLy+PoqIiEhMTcXFxUcttbW0xNzcnJSUFgLi4OLU8MDCQtm3bsnz5cho1anSXuiCEEKI2jN7ht27dmoCAAHx9fSktLWXEiBE4ODjg7+/P1KlT6dy5M6GhocyZM4erV6/y3HPP4evry/Hjx9m7dy/t27dXp2x0Oh2RkZF3vVNCCCGqMhr4AHq9Hr1eb7Dv5uDu2LEjMTExBuXPPvssJ06cqIcmCiGEqA/yTVshhNCIhzrwS0rL73cTHhgyFkIIY2o1pfOgamRmypgZVZd60KLNIfIxViHE7T3Ud/hCCCFqTwJfCCE0QgJfCCE0QgJfCCE0QgJfCCE0QgJfCCE0QgJfCCE0QgJfCCE0QgJfCCE0QgJfCCE0QgJfCCE0QgJfCCE0QgJfCCE0QgJfCCE0QgJfCCE0QgJfCCE0QgJfCCE0QgJfCCE0QgJfCCE0olaBHx8fz5AhQ3B1dSUqqurfkE1LS8Pb2xs3NzeCgoIoKyszKF++fDnh4eH102IhhBB/iNHAz87OJiwsjM2bNxMbG8vWrVtJT083OCYwMJB58+axZ88eFEUhOjoagIKCAmbPns26devuTuuFEELUmtHAT0pKwtnZGSsrKywsLHBzcyMhIUEtz8zMpLi4GEdHRwC8vb3V8r179/Lkk0/yyiuv1Fh/fn4+586dM/iXlZV1p/0SQghxi4bGDsjJycHa2lrd1ul0HDt2rMZya2trsrOzAfD09AS47XTOhg0biIiIqHvLhRBC1InRwK+oqMDExETdVhTFYNtYuTF+fn54eXkZ7MvKysLHx6fWdQghhDDOaODb2NiQnJysbufm5qLT6QzKc3Nz1e0LFy4YlBtjaWmJpaVlrY8XQgjxxxidw+/duzcHDx4kLy+PoqIiEhMTcXFxUcttbW0xNzcnJSUFgLi4OINyIYQQDwajgd+6dWsCAgLw9fXF09MTd3d3HBwc8Pf3JzU1FYDQ0FAWL17MoEGDKCwsxNfX9643XAghRN0YndIB0Ov16PV6g32RkZHqzx07diQmJqbG8998880/2DwhhBD1Rb5pK4QQGiGBL4QQGiGBL4QQGiGBL4QQGiGBL4QQGiGBL4QQGiGBL4QQGiGBL4QQGiGBL4QQGiGBL4QQGiGBL4QQGiGBL4QQGiGBL4QQGiGBL4QQGiGBL4QQGiGBL4QQGiGBL4QQGiGBL4QQGiGBL4QQGiGBL4QQGiGBL4QQGiGBL4QQGlGrwI+Pj2fIkCG4uroSFRVVpTwtLQ1vb2/c3NwICgqirKwMgN9//x0fHx8GDRrEG2+8wbVr1+q39UIIIWrNaOBnZ2cTFhbG5s2biY2NZevWraSnpxscExgYyLx589izZw+KohAdHQ3AwoULGTNmDAkJCXTq1IlVq1bdnV4IIYQwqqGxA5KSknB2dsbKygoANzc3EhISmDJlCgCZmZkUFxfj6OgIgLe3NytWrGDkyJF8//33rFy5Ut0/duxYAgMDDerPz88nPz/fYF9mZiYAWVlZRjtwvfCy0WO04Ny5c3dcR25BcT205OFXH2MJUHy5sF7qedjV13jmXZfrE24/npWZWV5eXm250cDPycnB2tpa3dbpdBw7dqzGcmtra7Kzs7l06RLNmjWjYcOGBvtvtWHDBiIiIqp9bB8fH2PNE/8z8F8r7ncT/jyiB97vFvypDPxYxrM+LRpofDxzc3Np27Ztlf1GA7+iogITExN1W1EUg+2aym89DqiyDeDn54eXl5fBvpKSEjIyMnjyyScxNTU11sT7JisrCx8fH6KiorCxsbnfzXnoyXjWLxnP+vOwjGV5eTm5ubl06tSp2nKjgW9jY0NycrK6nZubi06nMyjPzc1Vty9cuIBOp6Nly5YUFBRQXl6OqalplfMqWVpaYmlpWWV/u3btjDXtgWFjY4Odnd39bsafhoxn/ZLxrD8Pw1hWd2dfyeibtr179+bgwYPk5eVRVFREYmIiLi4uarmtrS3m5uakpKQAEBcXh4uLC2ZmZjg5ObFr1y4AYmNjDc4TQghxbxkN/NatWxMQEICvry+enp64u7vj4OCAv78/qampAISGhrJ48WIGDRpEYWEhvr6+AMyfP5/o6GiGDBlCcnIy06ZNu7u9EUIIUSOjUzoAer0evV5vsC8yMlL9uWPHjsTExFQ5z9bWlo0bN95hE4UQQtQH0wULFiy43414mJmbm9OzZ0/Mzc3vd1P+FGQ865eMZ/35M4yliaIoyv1uhBBCiLtP1tIRQgiNkMAXQgiNkMC/xdWrV3F3d1e/vpyUlIRer8fV1ZWwsLAazxs3bhxDhw7Fw8MDDw8Pjh49ChhfeO7PprpxqMsYZGdn07dvX4N9NZ1f2+fmYVTb67CmhQtvNWvWLFxdXdXn5V//+le91Psgqss1WJdr6Nbn5HbnP7ALSipC9eOPPyru7u7Kc889p2RkZChFRUVKv379lLNnzyqlpaXKhAkTlG+++abKeRUVFUrfvn2V0tJSg/1ZWVlK//79lUuXLinXrl1T9Hq9cvLkyXvVnXuuunGoyxh88803iqurq/LMM88YPb+2z83DqC7X4dChQ5UjR44oiqIos2bNUqKioqqt093dXcnOzjbYVx/1Pmjqcg3W5Rq69TlRlD82fq+++qry5ZdfKoqiKBEREUpISMhdG4vqyB3+TaKjo5k/f776jeBjx47Rtm1b2rRpQ8OGDdHr9SQkJFQ57/Tp0wBMmDCBYcOGsWnTJsBw4TkLCwt14bk/q+rGoS5jEBMTQ3h4uMG+ms6v7XPzMKrtdVjdwoXVjUFRURG///47s2fPRq/Xs2LFCioqKu643gdRXa7BulxDtz4nUPfnpbS0lO+//x43NzeD/fdSrT6HrxXvvfeewXZ1C8dVtwBcfn4+vXr1Yu7cuZSWluLr68tTTz1ldOG5P5vqxmHw4MG1HoNbwx5qXryvts/Nw6i212FNCxfe6sKFCzg7OzN//nyaN2/Oa6+9RkxMDBYWFndU74OoLtdgXa6hW58TqPvzUtsFJe8mucO/DWMLx1Xq2rUrISEhNG/enJYtWzJixAj2799f6/P/LKobhxUrVtzRGNQ0hloa2zsdgzZt2rBy5Up0Oh1NmjRh3Lhxt70+H+axrcs1eKf9rOv4VVf/vR5XCfzbuHVhuMoF4Pbu3au+IfThhx+SnJzMwYMH1eMURaFhw4Y1nv9nVd042NraVjsGn332mTqGn332WY111jSGWhrb2o5B5cKFqamp6tgGBQVx4sQJ9uzZox5n7Pqsqd6HQV2uwdr+/65JXcfv5gUlbz7+nrqn7xg8JPr3769kZGQoxcXFiouLi3LmzBmlrKxMmThxorJr164qx3/99deKp6enUlxcrBQUFCh6vV754Ycf1DeLLl68qBQWFirDhg1Tjh49eh96dG9UNw7Jycl1HoPq3rS99fzaPjcPs9pch0OHDlWSk5MVRVGUOXPmKJGRkVXqSUtLU1xcXJTLly8rJSUlyoQJE5T4+Pg7rvdBVJdr8I9cQ5XPiaIof2j8/P39lR07diiKoiirVq1SFixYcLeGoloS+NW4+UlNSkpS9Hq94urqqrz33ntKRUVFteeEhYUpgwYNUlxdXZX169er+3fs2KEMHTpUcXV1VVavXn1P2n8/VTcOdR2DmwP/dufX9rl5WNXmOkxLS1OGDx+uuLm5KdOnT1euX79ebV2bNm1SBg8erLz00kvKBx98oO6/03ofRHW5But6Dd38nNzu/JrG79y5c8rYsWOVwYMHKxMmTFAuX75c392/LVlaQQghNELm8IUQQiMk8IUQQiMk8IUQQiMk8IUQQiMk8IUQQiMk8EWt2dvbk5eXZ7AvISGBcePG3bXHLCsrY+3atXh4eDB06FAGDBjA5MmT1TVT7qWcnByeffZZLl++rO77xz/+gb29vUF7IiMjmTZtGqmpqUydOhW4se7KvHnzADh06BDu7u61ftyJEydy+vRpSkpK1HVY6iIiIoKvvvoKgA8//JDY2Ng61yH+HGQtHfFAe+eddyguLmb9+vW0aNECuLFc8vjx49m9ezdNmza9Z23R6XQ888wzJCcn8+KLLwLwzTff0L9/f77++mvatWsHwHfffYder6dz586sWLECgPT09D+0bkpZWRkZGRm0a9eOw4cP4+DgUOc6Dh06RPv27QF466236ny++POQwBf15tSpUwQFBVFSUoKiKIwYMQIfHx8APvroIxITE6moqMDW1pb58+fTunVrxo0bxyOPPMLp06d5+eWXDX5b+Omnnzh48CB79+6lSZMm6n69Xs+VK1e4evUqTZs2pVOnTgwcOJCff/6Z0NBQrl+/TkhICEVFRZiZmTFt2jRcXFzYtm0be/bs4ZNPPgEw2J45cybm5ub8/PPPXLx4kT59+jBnzhzMzMwM+uji4sKhQ4d48cUXOXfuHCUlJfj5+REeHs6kSZMoKSnhyJEjhISEcOjQIRYtWkRkZCQrVqygoKCAWbNm4enpSWFhIQEBAZw+fZrr168THByMk5OTwWP5+/tz+vRprl69ioeHB9nZ2TRt2pSoqChSU1O5fPkyGRkZvPDCC4wYMYJ3332Xa9eukZubS8eOHVm+fDkxMTH89NNPhISEYGpqyt69e+nQoQMTJ06kc+fOvPrqq3z77bfk5OQwadIkxowZQ3l5OSEhIXz99dc0b94cBwcHTp06xcaNG0lMTOSjjz7CxMQEU1NTZsyYQY8ePe7WJSXqmUzpiHqzdu1aBgwYwLZt21i9ejXJyclUVFQQGxvLL7/8wueff05cXBz9+vVjzpw56nmWlpbs2rWrytRQcnIy3bt3Nwj7SmPHjqV169YAlJaW0r9/f/bs2YOdnR1Tp04lKCiI+Ph4li5dSmBgIBkZGUbbf+zYMf75z3+ya9cuTp06xdatW6sc4+LiwuHDhwHYt28fL7zwAk5OTqSnp5OXl8fRo0dp3749rVq1Us957LHHmDp1Kk5OTixevBiArKwsxo8fT1xcHKNHj652pdDIyEgmTpzIhAkTiIuLw8nJifDwcPVFtLi4mJ07dxIYGEh0dDSenp5ER0eTmJjIuXPn+Oabb/Dx8aFTp07MmDGDl156yaD+kpISWrRowZYtW1ixYgWLFy/m+vXrfP755/z3v//lyy+/ZMuWLQZjFxISwvz589m2bRtvvfUWhw4dMjqu4sEhd/ii1qpb2a+iooIGDW7cN7z00ku88847HDt2jF69ejFnzhwaNGjAvn37SE1NZfjw4eo5RUVFah233tnW9JinT58mICAAgIKCAvWO9OY6jh07xhNPPEGXLl0A6NChA926dePw4cNGVyb08vJSp4g8PDzYu3cvY8eONTjG0dGR8+fPc/nyZfbt24e/vz9mZmY4Ozvz3XffcerUKfr163fbx4EbK1hWtrFjx4588cUX1R73888/q9NHJ0+eVKdmALp3767+HBgYyLfffktkZCRnzpwhJyeHwsJCo+0YOHAgAM899xwlJSUUFhayf/9+PDw8MDc3B+Cvf/0rGzduBGDo0KFMmTKFfv360adPH/z9/Y0+hnhwSOCLWmvRogWXL1+mZcuW6r6LFy9iZWUFoN5lJyUlcfDgQVauXMm2bduoqKgwCOeSkhKuXLmi1mFhYVHt43Xt2pU1a9ZQWlqKmZkZ7dq1Iy4uDoCZM2cavGhU1lFeXl4l2BVFoaysjEaNGnHzSiKlpaUGx5mamhqcU/lCdrOGDRvi7OzMv//9b9LS0tQXmn79+pGSksLPP//M7NmzaxpC1c1TRZVL597K39+fw4cP88MPPxASEkJ2djYjR45k1KhRBn0GmD59OuXl5QwePJgXXniB8+fPV1vnrSpDvXLMlP+tpHmzm8chICCA4cOH8+2337Jt2zb++c9/EhMTY/RxxINBpnRErbm4uLBx40YqKioAuHLlCtu3b1fvaP/+97+za9cuhg4dyvz582nWrBlnz56lb9++xMTEcPXqVeDGJ0VmzJhh9PG6dOlCz549mTFjhsGng9LT00lLSzMI6EqOjo6cPn1a/SMrJ0+e5Pvvv+f555+nZcuWnDx5kuvXr1NaWmqwZDDA7t27KSkp4fr162zfvp3+/fvXOA5r1qzh+eefV4O7X79+HDx4kPPnz/Pss89WOcfU1LTOfxd2+fLl2Nra8uWXXzJt2jRGjBhBXFycOqVzswMHDvC3v/2NIUOGAHD06FF1Gd66Pna/fv3YsWMHJSUllJWVsX37duDGG8gDBgygqKiIl19+mfnz53PixAlKSkrq1C9x/8gdvqi1oKAglixZgru7uxq2Hh4eeHl5ATB58mSCgoLYunUrpqamvPjii/To0QMnJyeys7MZNWoUJiYmPPbYYyxZsqRWj/nBBx8QFRXFq6++SllZGVeuXMHGxoYxY8bg7e1d5fiWLVvy4YcfsmjRIoqLizExMWHx4sU89dRTtGnThh49eqh/Aalnz56cOHFCPbdx48aMGTOG/Px83Nzc1CmoW7m4uBAUFMSECRPUfY8++igWFhY4OjpWO3Xk6OjIypUrmTJlSq0/xvrjjz/SrVs34Mb7Gbd7czQgIIC//e1vWFhY0KxZM3r06MHZs2cBGDBgAMuWLavyG01NvL29+fXXX/H09MTCwgI7OzuaNGlCw4YNmT17Nm+//TYNGzbExMSE999/n0aNGtWqXnH/yWqZQnBjiqjy0ytad+DAAS5evIiHhwcAwcHBmJubExgYeJ9bJu6UTOkIIQx06NCB2NhY9Ho9Q4cO5dKlS7z++uv3u1miHsgdvhBCaITc4QshhEZI4AshhEZI4AshhEZI4AshhEZI4AshhEZI4AshhEb8P3d8exd5BIQfAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "User coverage:  the fraction of users for which at least 1 items can be recommended well is 0.188\n",
      "Item coverage: the fraction of items that can be recommended to at least 1 users well is 0.353\n",
      "Catalog coverage: the fraction of items that are in the top-10 for at least 1 user is 0.353\n",
      "\n",
      "NDCG: 0.310\n"
     ]
    }
   ],
   "source": [
    "# Evaluation\n",
    "\n",
    "k_items = 10\n",
    "rating_pred = pd.DataFrame(index = rating_test.index, columns = ['dummy']).join(rating_pred_mfnn.pivot(index='movieId', columns='userId', values='rating'), how = 'left').drop('dummy', axis = 1)\n",
    "\n",
    "\n",
    "#recall\n",
    "user_recalls, avg_recall = recall_over_user(rating_test, rating_pred, k_items=k_items)\n",
    "print(f\"Mean Recall: {avg_recall:.3f}\")\n",
    "user_grouped = group(train_df, 'user')\n",
    "recall_dict = recall_over_user_group(rating_test, rating_pred, user_grouped=user_grouped, k_items=k_items)\n",
    "\n",
    "#coverage\n",
    "user_cov = cal_coverage(rating_test, rating_pred, k_items=k_items, threshold=1, coverage_type='user')\n",
    "movie_cov = cal_coverage(rating_test, rating_pred, k_items=k_items, threshold=1, coverage_type = 'item')\n",
    "catalog_cov = cal_coverage(rating_test, rating_pred, k_items=k_items, threshold=1, coverage_type = 'catalog')\n",
    "\n",
    "#ndcg\n",
    "ndcg_score = ndcg(rating_test, rating_pred)\n",
    "print()\n",
    "print(f\"NDCG: {ndcg_score:.3f}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<b>Result Interpretation</b>: \n",
    "\n",
    "As the result shows, using MF-NN hybrid model fails to beat the single MF model in terms of `recall@10`, `user coverage` and `NDCG`. The only improvement is `movie coverage`, increasing from `29.6%` to `35.3%`. It also takes more time to train, which is not worth the time clearly."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 5.1.2 MF-NN Model (MF embedding + movie tag embedding)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We try to add in some side information like movie tag relevance. So we process it into a 1128-dimension tag embeddings for each movie, and feed them into the Neural Netwerok as well, to see if it can further improve the resutls."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training begins.......\n",
      "training set created\n",
      "start training neural network......\n",
      "model training finished\n",
      "creating testing set for neural network.....\n",
      "prediction finished\n",
      "train_time:  699.2198009490967\n",
      "test_time:  128.2756917476654\n"
     ]
    }
   ],
   "source": [
    "#model initializing\n",
    "tuned_model_mfnn = mf_nn(max_iter = 15, regparam = 0.1, k = 30, rank = 10, movie_tag_embed=True) \n",
    "\n",
    "#training\n",
    "start = time.time()    \n",
    "tuned_model_mfnn.fit(train_df, df_tag)\n",
    "end = time.time()\n",
    "train_time = end - start\n",
    "\n",
    "#testing\n",
    "start = time.time()\n",
    "rating_pred_mfnn = tuned_model_mfnn.pred(test_df)\n",
    "\n",
    "end = time.time()\n",
    "test_time = end - start\n",
    "print(\"train_time: \", train_time)\n",
    "print(\"test_time: \", test_time)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_history(tuned_model_mfnn.history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZMAAAEXCAYAAABoPamvAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO3dfVyV9f3H8dcREDNk3sQRAufUnDZNSflpmIPKEhSOBuZKmeKc86ZNnRaFwiBsTitSWw7T1s85s5W3oI7QyqmpzdSVimm5TBQUPKjJjXJ//f7o53l4RDvIEZB8Px+P84Dre65zvp+PKG+v6zrne0yGYRiIiIg4oUlDFyAiIo2fwkRERJymMBEREacpTERExGkKExERcZrCREREnKYwkZuqS5cuWCwWhg4dyuOPP05ISAjDhg3j4MGDN32u7Oxs7r//fgBef/11Zs2aVW2ftWvX0rt3b4YOHcrQoUOxWCxMnDiRzMxM2z6/+c1v+O9///u9c40dO5Zz585d877Lj9+9ezfh4eE33MeqVatYsWIFAP/4xz9YsmTJDT/H9/nHP/7BkCFDGDx4MGFhYcTExHDq1KmbOseVrvezCA8PZ/fu3XU2rzQs14YuQH54li1bRuvWrW3bb731Fn/84x957733GqSegIAAFi9ebNvetWsX48aNY82aNfj6+vLmm286fI6dO3de977Ljz979myt6tu3bx+dO3cGYMSIEbV6jut56aWXOHLkCIsXL8bHx4eqqirWr1/Pk08+yapVq/D29r6p88ntS2EidaqiooLTp0/zox/9yDa2aNEiNm/eTFVVFb6+viQmJtK2bVusViuJiYkcO3aMJk2a8NRTTzF69Gg+//xzXnnlFcrKyrBarfTr148//elPta6pX79+PPbYY/zjH//g2Wef5ZFHHuG1116jY8eOzJgxg6ysLJo0aUK3bt2YNWsWcXFxAERHR7NkyRKioqLo0aMHX375JdOnT2fOnDm89tprAFy8eJEpU6aQlZWFp6cns2bNokOHDsTGxtK5c2d+/etfA9i2f/zjH7NlyxZ27txJs2bNOHfuHOfPnychIYGjR48ya9Ysvv32W0wmE2PHjuXxxx9n9+7dzJ8/n3bt2nH06FEqKipISkqid+/edn3m5uby7rvvsnXrVtuff5MmTXj88cfJzMxk8eLFDBgwgJdeeokNGzYAUFBQwIABA/jwww8pKSlh1qxZnD59mvLycsLCwpg4cSLZ2dlERUXRqVMncnJyWL58OWaz+YZ+Bn/+85/54IMPcHNzo1WrVsyZMwez2czXX3/N7Nmz+fbbb6msrGTUqFE88cQT7N69m9mzZ9O8eXOKi4t55513iIuLq/azatJEJ1saiv7k5aaLjo7GYrHQv39/QkJCAJgzZw4AqampfPXVV6xatYq0tDSCg4OJj48HICkpiZ/85CdkZGTw3nvvsXLlSrKysvj73//OlClTWLVqFf/85z/ZsmWL3Wmq2ujatStfffWV3dgHH3xAcXExaWlprF69GoCTJ0/aal+2bBk+Pj4AdO7cmffff5/HHnvM7jlOnz7NmDFjSEtLIzw8nOeee+5763jsscd45JFHGDNmDFFRUbbxiooKJk2axKhRo9iwYQNvvvkm8+bN47PPPgPgwIEDjB07ltTUVCIjI5k/f361596/fz8dO3a0C/LL+vXrx759+3jwwQcpLi62nYbcuHEjwcHB/OhHPyImJoZhw4axdu1aVq9eza5du0hPTwe+C6qnn36aTZs23XCQnD59mmXLlrFmzRrWrl3Lgw8+yIEDB6ioqGDKlCk888wzrF27lrfffpv//d//5fPPPwfg6NGjvPrqq2zYsIGPPvromj8raTg6MpGb7vJprkOHDjF+/Hj69u1LmzZtAPjXv/7FwYMHGTZsGABVVVVcunQJ+O70U0xMDAAtWrRg48aNAMydO5ft27fzxhtvcOzYMUpLS7l48SItW7Z0qs5mzZrZbffu3Zv58+czatQo+vXrR3R0NO3bt7/mYwMCAq453qVLF3r16gVAREQEL7zwAoWFhTdc2/HjxyktLWXgwIEAtG3bloEDB/Lxxx/Tt29f7r77bu69914Afvazn7Fu3bprPk9FRcU1x8vKyjCZTJhMJoYNG8a6deu47777WLt2Lc899xwXL15kz549XLhwwe6o68iRI/To0QNXV1f8/f2v+dwmk+ma41VVVbi4uNC2bVu6du1KREQEQUFBBAUFERgYyH//+19OnDjBzJkzbY8pKSnhiy++oFOnTvj4+ODr6wvc2M9K6ofCROpMt27dmDFjBrGxsdx77734+flRVVXFuHHjGDlyJPDdL7ULFy4A4OrqaveL6OTJk7Rq1YqxY8fSpUsXfv7znzNo0CD279+Ps0vKZWZm8tOf/tRurF27dnzwwQfs3r2bf//73/zqV79i1qxZPPLII9Ue37x582s+79WnWUwmk62vK2suLy//3voqKyur/VI2DMMWDlcG4dXPfZm/vz9ZWVlYrVa8vLzs7tu9e7ftxQtPPPEEERERDB8+nMLCQvr06UNRURGGYfDuu+9yxx13AHDu3Dnc3d05f/48TZs2xdX12r8+WrVqxbFjx6qNnz17lpYtW9KkSRPefvttDh48yCeffMKf/vQnfv7znzN06FBatGhBWlqa7TH5+fm0aNGCzz//3O7P/EZ+VlI/dJpL6lR4eDg9evSwnSrq378/q1evpqioCIDXXnvNdiooMDCQNWvWAFBYWEh0dDTHjx/n4MGDPPvsswwcOJDc3FxOnDhBVVVVrWvatm0bW7du5cknn7Qbf+edd5gxYwb9+/cnJiaG/v3788UXXwDg4uJy3f/lX+nLL7/k8OHDALz33nv07t2bO+64g1atWtlOzeXl5fHpp5/aHnOt5+7YsSOurq5s3rzZ9phNmzbRr1+/GvfZtm1bRo0axfTp08nLy7ONr1mzhs2bN/Ob3/zGtl+PHj1ISEjgiSeeAMDDwwN/f3+WLl0KfHctZcSIEXz00UcO5+3Xrx+ffPIJX3/9tW0sNTUVDw8POnbsyJEjRwgPD6dTp05MmDCBMWPGcPDgQTp06ECzZs1sYXL69GnCw8OveUrz+35W0jB0ZCJ17g9/+ANDhgzh448/Zvjw4eTl5fGLX/wCk8mEj48Pc+fOBSAhIYEXXngBi8WCYRhMmDCB7t27M378eCIiImjevDlt27alV69eZGVl0a5duxrNv3fvXoYOHQp89794s9nMW2+9Ve1/648//jiffvopgwcP5o477sDHx4dRo0YBEBoayqhRo3j99de/d66OHTuycOFCTp48SZs2bWy9jRo1imeffZaQkBD8/Px44IEHbI8JCgqy7XeZm5sbKSkp/PGPf+T111+nsrKS3/72tzzwwAM39PLaZ555hlWrVjFp0iTKysooKyvjvvvu491337WdMgIYPnw4U6dOZdGiRbax5ORkXnzxRSwWC2VlZYSHhzNkyBCys7O/d84OHTrw4osv8txzz1FRUUFZWRl+fn68+eabNGnShK5duzJo0CCGDRtG8+bNadasGfHx8TRt2pSUlBRmz57NX//6VyoqKpg6dSq9e/eu1vP3/aykYZi0BL2IiDhLp7lERMRpChMREXGawkRERJymMBEREafddq/mKikpITMzEy8vL1xcXBq6HBGRRqGyshKr1Ur37t2rveEXbsMwyczMtFu2QkREam7FihXXXAHitguTy+8tWLFihVZMFRGpodzcXKKioqq9P+uy2y5MLp/a8vb2xs/Pr4GrERFpXK53eUAX4EVExGkKExERcZrCREREnKYwERERpylMRETEaQoTERFxmsJEREScpjCphbLyyttqXhERR267Ny3eDE3dXLA8k+Z4x5tsw6tD631OEZGa0JGJiIg4TWEiIiJOU5iIiIjT6jRMioqKCA8PJzs7G4DPPvuMX/ziF4SFhTF9+nTKysoAOHz4MJGRkYSEhBAXF0dFRQUAp06dIioqitDQUCZNmkRxcTEABQUFjB8/nkGDBhEVFYXVaq3LNkRExIE6C5P9+/czYsQIjh8/DnwXLJMnT2bWrFn885//BGD16tUAxMTEkJCQwKZNmzAMg5UrVwKQlJTEyJEjycjIoHv37qSkpACwYMECAgICeP/99xk+fDizZ8+uqzZERKQG6ixMVq5cSWJiImazGYCdO3fi7+9P165dAYiPj+exxx4jJyeHkpIS/P39AYiMjCQjI4Py8nL27NlDSEiI3TjA1q1bsVgsAISHh7N9+3bKy8vrqhUREXGgzl4afPXRQlZWFs2bN2fatGkcO3aMXr16ERsbyxdffGH3YSteXl7k5eVx/vx5PDw8cHV1tRsHOHPmjO0xrq6ueHh4cO7cOdq2bWs3Z0FBAQUFBXZjubm5N71XEZHbXb29z6SyspIdO3bw3nvvcffddxMXF8eSJUvo168fJpPJtp9hGJhMJtvXK129feVjmjSpfpC1bNkyFi5ceHMbERGRauotTO666y569uxJu3btABg0aBBvv/02kZGRdhfQ8/PzMZvNtG7dmsLCQiorK3FxccFqtdpOmZnNZvLz8/H29qaiooLi4mJatmxZbc7o6GgiIiLsxi5/9KSIiNw89fbS4P79+3Po0CFOnz4NwL/+9S+6deuGr68v7u7u7Nu3D4C0tDSCgoJwc3MjICCA9PR0AFJTUwkKCgIgODiY1NRUANLT0wkICMDNza3anJ6envj5+dnd9LnvIiI3X70dmfj4+DBr1iwmTpxIaWkp9957L88//zwAycnJxMfHU1RURLdu3Rg9ejQAiYmJxMbGsmjRInx8fJg3bx4AU6dOJTY2lrCwMFq0aEFycnJ9tSEiItdQ52GyZcsW2/cPPfQQDz30ULV9unbtanuZ8JV8fX1Zvnx5tfGWLVvyxhtv3NQ6RUSk9vQOeBERcZrCREREnKYwERERpylMRETEaQoTERFxmsJEREScpjARERGnKUxERMRpChMREXGawkRERJymMBEREacpTERExGkKExERcZrCREREnKYwERERpylMRETEaXUaJkVFRYSHh5OdnW03/vbbbzNq1Cjb9uHDh4mMjCQkJIS4uDgqKioAOHXqFFFRUYSGhjJp0iSKi4sBKCgoYPz48QwaNIioqCi7z5AXEZH6V2dhsn//fkaMGMHx48ftxv/73/+yZMkSu7GYmBgSEhLYtGkThmGwcuVKAJKSkhg5ciQZGRl0796dlJQUABYsWEBAQADvv/8+w4cPZ/bs2XXVhoiI1ECdhcnKlStJTEzEbDbbxsrKykhISGDKlCm2sZycHEpKSvD39wcgMjKSjIwMysvL2bNnDyEhIXbjAFu3bsVisQAQHh7O9u3bKS8vr6tWRETEgTr7DPhrHS28+uqrDBs2DD8/P9vYmTNn8PLysm17eXmRl5fH+fPn8fDwwNXV1W786se4urri4eHBuXPnaNu2rd18BQUFFBQU2I3l5ubenAZFRMSmzsLkajt37uT06dPMmDGD3bt328arqqowmUy2bcMwMJlMtq9Xunr7ysc0aVL9IGvZsmUsXLjwJnUgIiLXU29hsnHjRo4ePcrQoUO5ePEi+fn5/P73vycmJsbuAnp+fj5ms5nWrVtTWFhIZWUlLi4uWK1W2ykzs9lMfn4+3t7eVFRUUFxcTMuWLavNGR0dTUREhN1Ybm4uUVFRddusiMhtpt7CZM6cObbvd+/ezcKFC1mwYAEA7u7u7Nu3j969e5OWlkZQUBBubm4EBASQnp6OxWIhNTWVoKAgAIKDg0lNTWXixImkp6cTEBCAm5tbtTk9PT3x9PSsnwZFRG5jt8T7TJKTk5kzZw6hoaFcvHiR0aNHA5CYmMjKlSsZPHgwe/fu5fe//z0AU6dO5fPPPycsLIx33nmHhISEhixfROS2V+dHJlu2bKk21rdvX/r27Wvb7tq1K6tXr662n6+vL8uXL6823rJlS954442bW6iIiNTaLXFkIiIijZvCREREnKYwERERpylMRETEaQoTERFxmsJEREScpjARERGnKUxERMRpChMREXGawkRERJymMBEREacpTERExGkKExERcZrCREREnKYwERERpylMRETEaXUaJkVFRYSHh5OdnQ3Ae++9R3h4OBaLhRkzZlBWVgbA4cOHiYyMJCQkhLi4OCoqKgA4deoUUVFRhIaGMmnSJIqLiwEoKChg/PjxDBo0iKioKLvPkBcRkfpXozA5efIkAFu3buUvf/kLhYWFDh+zf/9+RowYwfHjxwH45ptveOutt3j33XdZv349VVVVvPPOOwDExMSQkJDApk2bMAyDlStXApCUlMTIkSPJyMige/fupKSkALBgwQICAgJ4//33GT58OLNnz77hxkVE5OZxGCYJCQm8+eabfP3118THx5Odnc3MmTMdPvHKlStJTEzEbDYD0LRpUxITE/Hw8MBkMvHTn/6UU6dOkZOTQ0lJCf7+/gBERkaSkZFBeXk5e/bsISQkxG4cvgs1i8UCQHh4ONu3b6e8vLxaDQUFBWRnZ9vdcnNza/hHIyIiNeXwM+AzMzNZvXo1S5YsISIigmeeeYbIyEiHT3z10YKvry++vr4AnDt3jhUrVjBnzhzOnDmDl5eXbT8vLy/y8vI4f/48Hh4euLq62o0Ddo9xdXXFw8ODc+fO0bZtW7s5ly1bxsKFCx3WKiIiznEYJoZh0KRJE3bu3MnEiRMBKCkpqfWEeXl5jBs3jmHDhtG3b1/27duHyWSym89kMtm+Xunq7atrvFp0dDQRERF2Y7m5uURFRdW6fhERqc5hmPz4xz/mN7/5DdnZ2fTp04dnnnmGrl271mqyr7/+mnHjxjFq1CjGjh0LgLe3t90F9Pz8fMxmM61bt6awsJDKykpcXFywWq22U2Zms5n8/Hy8vb2pqKiguLiYli1bVpvP09MTT0/PWtUqIiI15/CayezZswkPD2f58uW4ubkREBBQqwveRUVF/PrXv2bq1Km2IIHvTn+5u7uzb98+ANLS0ggKCrLNlZ6eDkBqaipBQUEABAcHk5qaCkB6ejoBAQG4ubndcE0iInJzODwyiYqKIi0tzbY9YsSIWk20evVq8vPzWbp0KUuXLgXgkUceYerUqSQnJxMfH09RURHdunVj9OjRACQmJhIbG8uiRYvw8fFh3rx5AEydOpXY2FjCwsJo0aIFycnJtapJRERuDodhcscdd5Cbm4u3t3etJtiyZQsAY8aMYcyYMdfcp2vXrqxevbrauK+vL8uXL6823rJlS954441a1SMiIjefwzC5dOkSAwYMwNvbm+bNm9vGN2zYUKeFiYhI4+EwTOLi4uqjDhERacQcXoDv06cPzZo149ixY/j7++Pm5kafPn3qozYREWkkHIbJ2rVrmTFjBn/9618pLCzk6aefti13IiIiAjUIk+XLl/Pee+/h4eFBmzZtWLt2LcuWLauP2kREpJFwGCZNmjTBw8PDtu3j44OLi0udFiUiIo2LwzBp2bIlhw8fti1lsn79en70ox/VeWEiItJ4OHw118yZM5k6dSonTpygf//+uLu725aCFxERgRqESadOnUhLS+P48eNUVlbSoUMHLV0iIiJ2HJ7mys/PZ+vWrXTq1In169czbtw4jhw5Uh+1iYhII+EwTGJjYzl58iSffPIJ27dvZ+jQofzxj3+sj9pERKSRcBgm3377LWPGjGH79u2Eh4cTGRnJpUuX6qM2ERFpJByGSXl5OeXl5Xz88cf069ePS5cucfHixfqoTUREGgmHYTJgwAACAwNp1aoV3bt3Z/jw4YSHh9dHbSIi0kg4fDXXlClT+MUvfmH7fPXk5ORaf9KiiIj8MDkMk82bN1cbO3HiBAMHDqyTgkREpPFxGCZXfjhVeXk5X375JX369KlRmBQVFfHUU0/xxhtv4Ofnx65du5gzZw6lpaUMGjSIadOmAXD48GHi4uIoLi4mICCApKQkXF1dOXXqFDExMZw9e5YOHTqQnJzMnXfeSUFBAc8++ywnT56kdevWLFiwAC8vLyf+GERExBk1Wujx8u3dd99l1apVNGvWzOET79+/nxEjRnD8+HEASkpKmDlzJikpKaSnp5OZmcm2bdsAiImJISEhgU2bNmEYhm1V4qSkJEaOHElGRgbdu3e3vfN+wYIFBAQE8P777zN8+PBafSa9iIjcPA7D5Gr33HMPx44dc7jfypUrSUxMxGw2A3DgwAHat29Pu3btcHV1xWKxkJGRQU5ODiUlJfj7+wMQGRlJRkYG5eXl7Nmzh5CQELtxgK1bt2KxWAAIDw9n+/btlJeXV6uhoKCA7Oxsu1tubu6NtiwiIg7c0DUTwzDIzMzE1dXhw6odLZw5c8buVJTZbCYvL6/auJeXF3l5eZw/fx4PDw/bXJfHr34uV1dXPDw8OHfunO1FApctW7aMhQsXOqxVREScc0PXTEwmE61bt2bu3Lk3PFFVVZVt5WH4LphMJtN1xy9/vdLV21c+pkmT6gdZ0dHRRERE2I3l5uYSFRV1w/WLiMj13VCYOMPb2xur1WrbtlqtmM3mauP5+fmYzWZat25NYWEhlZWVuLi42PaH745q8vPz8fb2pqKiguLiYlq2bFltTk9PTzw9PW9K/SIicn03fM2ktnr27Mk333xDVlYWlZWVbNy4kaCgIHx9fXF3d2ffvn0ApKWlERQUhJubGwEBAaSnpwOQmppKUFAQAMHBwaSmpgKQnp5OQECAVjIWEWlAji9+3CTu7u7MnTuXyZMnU1paSnBwMKGhocB3b4SMj4+nqKiIbt26MXr0aAASExOJjY1l0aJF+Pj4MG/ePACmTp1KbGwsYWFhtGjRguTk5PpqQ0RErsFkGIZxrTs+/PBDHn30UcrKymjatGl911VnsrOzGTBgAB999BF+fn61fh7LM2k3saqa2fDq0HqfU0QEHP/uvO5prtdeew2AJ598su6qExGRH4Trnua68847CQkJIS8vz/aejitt2LChTgsTEZHG47ph8te//tW2zMkf/vCH+qxJREQameuGiYeHB//zP//D4sWLMZvNHDp0iIqKCnr06IGHh0d91igiIrc4h6/mKiwsZNSoUdx1111UVlaSl5fHG2+8Qa9eveqjPhERaQQchslLL71EcnIyDzzwAACffPIJc+fOtS3GKCIi4vBNi8XFxbYgAQgMDNRnwIuIiB2HYWIymcjJybFtZ2dn4+LiUqdFiYhI4+LwNNdvf/tbnnzySQIDAzGZTOzYsYPExMT6qE1ERBoJh2Hy6KOP0rFjR/79739TVVXFhAkT6NSpU33UJiIijUSN1ubq2LEjHTt2rOtaRESkkaq3VYNFROSHS2EiIiJOcxgmzz33XH3UISIijZjDMDl8+DDXWaVeREQEqMEFeLPZTFhYGD179uTOO++0jcfHx9d60rS0NJYsWQJAUFAQzz//PLt27WLOnDmUlpYyaNAgpk2bBmBbbLK4uJiAgACSkpJwdXXl1KlTxMTEcPbsWTp06EBycrJdfSIiUn8cHpncf//9DB48GF9fX1q2bGm71dalS5eYPXs2y5cvJy0tjb1797JlyxZmzpxJSkoK6enpZGZmsm3bNgBiYmJISEhg06ZNGIZhW8YlKSmJkSNHkpGRQffu3UlJSal1TSIi4hyHRya/+93vKCkpISsri86dO1NaWsodd9xR6wkrKyupqqri0qVLNG/enIqKCjw8PGjfvj3t2rUDwGKxkJGRwT333ENJSQn+/v4AREZG8uc//5nhw4ezZ88e/vKXv9jGf/nLXxITE1PrukREpPYcHpns37+fRx99lAkTJnDmzBkeeugh/vOf/9R6Qg8PD6ZOncqgQYMIDg7G19eXM2fO4OXlZdvHbDaTl5dXbdzLy4u8vDzOnz+Ph4cHrq6uduNXKygoIDs72+6Wm5tb69pFROTaarRq8N/+9jeeffZZvL29efnll5k9ezZr1qyp1YRHjhxhzZo1/Otf/6JFixY8++yzHD9+HJPJZNvHMAxMJhNVVVXXHL/89UpXbwMsW7aMhQsX1qpOERGpOYdhUlJSwj333GPbDg4OZv78+bWecMeOHQQGBtKmTRvgu1NUb731lt3ikVarFbPZjLe3N1ar1Taen5+P2WymdevWFBYWUllZiYuLi23/q0VHRxMREWE3lpubS1RUVK3rFxGR6hye5nJ1deXChQu2//kfO3bMqQm7du3Krl27uHjxIoZhsGXLFnr27Mk333xDVlYWlZWVbNy4kaCgIHx9fXF3d2ffvn3Ad68CCwoKws3NjYCAANLT0wFITU0lKCio2lyenp74+fnZ3by9vZ2qX0REqnN4ZDJp0iR++ctfYrVamT59Ojt37mTWrFm1nrB///588cUXREZG4ubmxn333cfkyZN58MEHmTx5MqWlpQQHBxMaGgpAcnIy8fHxFBUV0a1bN0aPHg1AYmIisbGxLFq0CB8fH+bNm1frmkRExDkmowbvSMzKymLnzp1UVVURGBjYqFcNzs7OZsCAAXz00Uf4+fnV+nksz6TdxKpqZsOrQ+t9ThERcPy7s0Zrc1VUVFBVVYWrq6vtFVQiIiKXOQyTNWvWMHr0aA4ePMjevXuJiopi06ZN9VGbiIg0Eg4PM/72t7+xbt0626ulTp06xYQJEwgJCanz4kREpHFweGTi5uZm97Lbu+++Gzc3tzotSkREGpfrHpkcOnQIgC5dujBr1iyefPJJXFxcWLt2Lb169aq3AkVE5NZ33TCZPHmy3fbWrVtt35tMJqdWDRYRkR+W64bJli1b6rMOERFpxBxegLdaraxbt45vv/3WblyfwCgiIpc5vAA/adIkDhw4gGEYdjcREZHLHB6ZlJeXa+VdERH5Xg6PTLp168ZXX31VH7WIiEgj5fDIpFevXjz++ON4eXnZLaXy0Ucf1WlhIiLSeDgMk7feeovk5GR+/OMf10c9IiLSCDkME09PTwYPHlwftYiISCPlMEweeOABXnrpJQYOHEjTpk1t4926davTwkREpPFwGCYbNmwAsFsp2GQy6ZqJiIjYOAyTungn/JYtW1i4cCGXLl3iwQcfJD4+nl27djFnzhxKS0sZNGgQ06ZNA+Dw4cPExcVRXFxMQEAASUlJuJTF7vYAABJ3SURBVLq6curUKWJiYjh79iwdOnQgOTmZO++886bXKiIijjl8afDSpUuveautkydPkpiYSEpKCuvXr+eLL75g27ZtzJw5k5SUFNLT08nMzGTbtm0AxMTEkJCQwKZNmzAMg5UrVwKQlJTEyJEjycjIoHv37qSkpNS6JhERcY7DMPnqq69st8zMTJYuXcqRI0dqPeEHH3zA4MGD8fb2xs3Njfnz53PHHXfQvn172rVrh6urKxaLhYyMDHJycigpKcHf3x+AyMhIMjIyKC8vZ8+ePbbPVLk8LiIiDcPhaa45c+bYbefl5REXF1frCbOysnBzc2PixImcPn2ahx56iM6dO+Pl5WXbx2w2k5eXx5kzZ+zGvby8yMvL4/z583h4eNje93J5/GoFBQUUFBTYjeXm5ta6dhERubYb/kD3tm3bkpOTU+sJKysr2bt3L8uXL6d58+ZMmjSJZs2aYTKZbPsYhoHJZKKqquqa45e/XunqbYBly5ZpKRgRkXrgMEyuvD5iGAaZmZm0adOm1hPeddddBAYG0rp1awAeffRRMjIycHFxse1jtVoxm814e3tjtVpt4/n5+ZjNZlq3bk1hYSGVlZW4uLjY9r9adHQ0ERERdmO5ublERUXVun4REanuhq6ZHD16FB8fH5KTk2s94cMPP8yOHTsoKCigsrKSjz/+mNDQUL755huysrKorKxk48aNBAUF4evri7u7O/v27QMgLS2NoKAg3NzcCAgIID09HYDU1FSCgoKqzeXp6Ymfn5/dzdvbu9a1i4jItd3wNRNn9ezZk3HjxjFy5EjKy8t58MEHGTFiBB07dmTy5MmUlpYSHBxMaGgoAMnJycTHx1NUVES3bt0YPXo0AImJicTGxrJo0SJ8fHyYN2/eTa1TRERqzmGYfPbZZ8ybN48LFy7YfY7J5Tcz1sYTTzzBE088YTcWGBjI+vXrq+3btWtXVq9eXW3c19eX5cuX17oGERG5eRyGSUJCApGRkfzsZz+75kVuERERh2Hi6urKr371q/qoRUREGimHF+A7d+7Ml19+WR+1iIhII+XwyOTkyZMMGzaMu+++G3d3d9u4M9dMRETkh8VhmFxecFFEROR6HIZJnz596qMOERFpxBxeMxEREXFEYSIiIk5TmIiIiNMUJiIi4jSFiYiIOE1hIiIiTlOYiIiI0xQmIiLiNIWJiIg4TWEiIiJOa9Aweemll4iNjQVg165dWCwWBg4cyPz58237HD58mMjISEJCQoiLi6OiogKAU6dOERUVRWhoKJMmTaK4uLhBehARkQYMk08++YR169YBUFJSwsyZM0lJSSE9PZ3MzEy2bdsGQExMDAkJCWzatAnDMFi5ciUASUlJjBw5koyMDLp3705KSkpDtSIicttrkDD59ttvmT9/PhMnTgTgwIEDtG/fnnbt2uHq6orFYiEjI4OcnBxKSkrw9/cHIDIykoyMDMrLy9mzZw8hISF24yIi0jAcrhpcFxISEpg2bRqnT58G4MyZM3h5ednuN5vN5OXlVRv38vIiLy+P8+fP4+Hhgaurq9341QoKCigoKLAby83NrYuWRERua/UeJqtWrcLHx4fAwEDWrl0LQFVVld3nyxuGgclkuu745a9Xutbn0y9btoyFCxfWUSciInJZvYdJeno6VquVoUOHcuHCBS5evEhOTg4uLi62faxWK2azGW9vb6xWq208Pz8fs9lM69atKSwspLKyEhcXF9v+V4uOjiYiIsJuLDc3l6ioqLprUETkNlTvYbJ06VLb92vXruXTTz8lKSmJgQMHkpWVhZ+fHxs3bmTYsGH4+vri7u7Ovn376N27N2lpaQQFBeHm5kZAQADp6elYLBZSU1MJCgqqNpenpyeenp712Z6IyG2pQa6ZXM3d3Z25c+cyefJkSktLCQ4OJjQ0FIDk5GTi4+MpKiqiW7dujB49GoDExERiY2NZtGgRPj4+zJs3ryFbEBG5rTVomERGRhIZGQlAYGAg69evr7ZP165dWb16dbVxX19fli9fXuc1ioiIY3oHvIiIOE1hIiIiTlOYiIiI0xQmIiLiNIWJiIg4TWEiIiJOU5iIiIjTFCYiIuI0hYmIiDhNYSIiIk5TmIiIiNMUJiIi4jSFiYiIOE1hIiIiTlOYiIiI0xQmIiLitAYJk4ULFxIWFkZYWBgvv/wyALt27cJisTBw4EDmz59v2/fw4cNERkYSEhJCXFwcFRUVAJw6dYqoqChCQ0OZNGkSxcXFDdGKiIjQAGGya9cuduzYwbp160hNTeXQoUNs3LiRmTNnkpKSQnp6OpmZmWzbtg2AmJgYEhIS2LRpE4ZhsHLlSgCSkpIYOXIkGRkZdO/enZSUlPpuRURE/l+9h4mXlxexsbE0bdoUNzc3OnXqxPHjx2nfvj3t2rXD1dUVi8VCRkYGOTk5lJSU4O/vD3z3Mb8ZGRmUl5ezZ88eQkJC7MavVlBQQHZ2tt0tNze3XvsVEbkd1PtnwHfu3Nn2/fHjx3n//ff55S9/iZeXl23cbDaTl5fHmTNn7Ma9vLzIy8vj/PnzeHh44Orqajd+tWXLlrFw4cI67EZERKABwuSyo0ePMmHCBJ577jlcXFw4fvy47T7DMDCZTFRVVWEymaqNX/56pau3AaKjo4mIiLAby83NJSoq6uY2IyJym2uQMNm3bx9Tpkxh5syZhIWF8emnn2K1Wm33W61WzGYz3t7eduP5+fmYzWZat25NYWEhlZWVuLi42Pa/mqenJ56envXSk4jI7azer5mcPn2a3/72tyQnJxMWFgZAz549+eabb8jKyqKyspKNGzcSFBSEr68v7u7u7Nu3D4C0tDSCgoJwc3MjICCA9PR0AFJTUwkKCqrvVkRE5P/V+5HJW2+9RWlpKXPnzrWNPfXUU8ydO5fJkydTWlpKcHAwoaGhACQnJxMfH09RURHdunVj9OjRACQmJhIbG8uiRYvw8fFh3rx59d2KiIj8P5NhGEZDF1GfsrOzGTBgAB999BF+fn61fh7LM2k3saqa2fDq0HqfU0QEHP/u1DvgG5Gy8srbal4RaTwa7NVccuOaurnoiEhEbkk6MhEREacpTERExGkKExERcZrCREREnKYwERERpylMRETEaQoTERFxmsJEREScpjARhxryHfB6971I46B3wItDDfXOe9C770UaCx2ZyC1N65GJNA46MpFbmtYjE2kcdGQiIiJOU5iIXINedCByYxr1aa4NGzawaNEiKioqiI6OJioqqqFLkh8IvehA5MY02jDJy8tj/vz5rF27lqZNm/LUU0/Rt29f7rnnnoYuTUTkttNow2TXrl088MADtGzZEoCQkBAyMjL43e9+Z9unoKCAgoICu8fl5OQAkJub69T85RfPOfX42sjOzr6t5m3IuRu6Z5FbzeXfmZWV1z4N22g/A37x4sVcvHiRadOmAbBq1SoOHDjAiy++aNvn9ddfZ+HChQ1VoojID86KFSsICAioNt5oj0yqqqowmUy2bcMw7LYBoqOjiYiIsBsrKyvj5MmT/OQnP8HFxeWG583NzSUqKooVK1bg7e1du+JvAT+UPkC93Ip+KH2AermssrISq9VK9+7dr3l/ow0Tb29v9u7da9u2Wq2YzWa7fTw9PfH09Kz22I4dO96U+f38/Jx+nob2Q+kD1Mut6IfSB6gXgPbt21/3vkb70uB+/frxySefcO7cOS5dusTmzZsJCgpq6LJERG5LjfbIpG3btkybNo3Ro0dTXl7OE088QY8ePRq6LBGR21KjDRMAi8WCxWJp6DJERG57Li+88MILDV1EY+Pu7k7fvn1xd3dv6FKc8kPpA9TLreiH0geol5potC8NFhGRW0ejvQAvIiK3DoWJiIg4TWFyHRs2bGDw4MEMHDiQFStWVLv/8OHDREZGEhISQlxcHBUVFQ1QZc046uWy5557jrVr19ZjZTfOUS8ffvghQ4cOZciQITz99NNcuHChAaqsGUe9fPDBB1gsFsLCwoiNjaWsrKwBqnSspn+/tm7dyiOPPFKPld04R70sXLiQhx9+mKFDhzJ06NDv7behOerl2LFjjBo1iiFDhvDrX//a+X8rhlSTm5trPPzww8b58+eN4uJiw2KxGEePHrXbJywszPjss88MwzCMGTNmGCtWrGiIUh2qSS+5ubnGhAkTjB49ehhr1qxpoEodc9RLYWGh8eCDDxq5ubmGYRjGggULjBdffLGhyv1ejnopLi42+vfvb1itVsMwDOP3v/+98e677zZUuddVk79fhmEYVqvVCA0NNR5++OEGqLJmatLLhAkTjP/85z8NVGHNOeqlqqrKGDhwoLFt2zbDMAzjlVdeMV5++WWn5tSRyTVcuYhk8+bNbYtIXpaTk0NJSQn+/v4AREZG2t1/K3HUC3z3P5gBAwYwaNCgBqqyZhz1Ul5eTmJiIm3btgWgS5cunD59uqHK/V6OemnevDlbtmzhrrvu4tKlS5w9e/aaqzk0tJr8/QKIj4+3W4T1VlSTXjIzM1m8eDEWi4VZs2ZRWlraQNV+P0e9HDp0iObNm9ve6D1x4kSnP8JDYXINZ86cwcvLy7ZtNpvJy8u77v1eXl52999KHPUCMG7cOIYPH17fpd0wR720atWKxx57DICSkhKWLFnCo48+Wu911kRNfi5ubm5s27aNhx56iPPnz9O/f//6LtOhmvTx97//nZ/97Gf07Nmzvsu7IY56KS4u5t577yUmJoZ169ZRUFBASkpKQ5TqkKNeTpw4wV133cXMmTOJiIggMTGR5s2bOzWnwuQaHC0iWZNFJm8VjalWR2raS2FhIePHj6dr167VFvq8VdS0l+DgYHbv3s3DDz/MrfiWMEd9fPXVV2zevJmnn366Icq7IY56ufPOO3nzzTfp1KkTrq6ujB07lm3btjVEqQ456qWiooJPP/2UESNGsG7dOtq1a8fcuXOdmlNhcg3e3t5YrVbb9tWLSF59f35+frVFJm8VjnppTGrSy5kzZxg5ciRdunRh9uzZ9V1ijTnq5dtvv2XHjh22bYvFwpdfflmvNdaEoz4yMjKwWq0MGzaM8ePH234+tyJHvZw6dYrVq1fbtg3DwNX11lxExFEvXl5etG/fnvvuuw+A8PBwDhw44NScCpNrcLSIpK+vL+7u7uzbtw+AtLS0W3aRyR/SgpiOeqmsrGTixIkMGjSIuLi4W/oIzFEvhmEQExPDqVOngO9+Kffq1auhyr0uR31MmTKFTZs2kZaWxpIlSzCbzbzzzjsNWPH1OeqlWbNmvPLKK5w8eRLDMFixYoXttOqtxlEv999/P+fOnePIkSMAbNmyhW7dujk3qVOX73/A1q9fb4SFhRkDBw40lixZYhiGYYwbN844cOCAYRiGcfjwYWPYsGFGSEiIMX36dKO0tLQhy/1ejnq57Pnnn7+lX81lGN/fy+bNm40uXboYQ4YMsd1mzpzZwBVfn6OfywcffGCEh4cbFovFmDZtmlFQUNCQ5V5XTf9+nTx58pZ+NZdhOO4lIyPDdn9sbGyj/nf/+eefG8OGDTMGDx5sjB071sjPz3dqPi2nIiIiTtNpLhERcZrCREREnKYwERERpylMRETEaQoTERFxmsJEpBF65JFHOHjwILt37yY8PLyhyxFRmIiIiPNuzbUARH4Adu/ezezZs2nevDnFxcVMnTqVxYsXU15eTrNmzXj++ee5//77qaio4JVXXmHr1q24uLhw//33k5iYSEFBAQkJCZw9exar1Yqvry8LFiygTZs2Dd2aSDUKE5E6dPToUT788EPKy8uZPHkyf//732nVqhVHjx7lV7/6FZs3b2b16tUcOnSItLQ0mjZtyvTp00lPT+fChQv4+/szfvx4DMNg/PjxpKWlMXbs2IZuS6QahYlIHfLx8cHX15cVK1Zw5swZxowZY7vPZDJx4sQJdu3axdChQ2nWrBkACxYssO2zd+9eli5dyvHjxzl69Ogtv4y73L4UJiJ16PJnRFRVVREYGGgXFKdPn8ZsNldbeTY/P5+qqiqWLVvGgQMHGDZsGH379qWiogKtfiS3Kl2AF6kHgYGB7Ny5k6+//hqAbdu2MWTIEEpKSggMDGTjxo2UlZVRVVXFCy+8wD//+U927NhBdHQ0jz/+OG3atGHXrl1UVlY2cCci16YjE5F6cM899zBr1iymT59u+xyMRYsWceedd/LUU0+Rk5NDZGQkhmHQp08fRo0aha+vLy+//DKvvfYabm5u9OrVixMnTjR0KyLXpFWDRUTEaTrNJSIiTlOYiIiI0xQmIiLiNIWJiIg4TWEiIiJOU5iIiIjTFCYiIuI0hYmIiDjt/wBAllZ7l1izlAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean Recall: 0.025\n",
      "Recall for each user group:  {'10-50': 0.016, '50-100': 0.0184, '100-500': 0.0215, '500-1000': 0.0}\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "User coverage:  the fraction of users for which at least 1 items can be recommended well is 0.166\n",
      "Item coverage: the fraction of items that can be recommended to at least 1 users well is 0.307\n",
      "Catalog coverage: the fraction of items that are in the top-10 for at least 1 user is 0.307\n",
      "\n",
      "NDCG: 0.309\n"
     ]
    }
   ],
   "source": [
    "# Evaluation\n",
    "\n",
    "k_items = 10\n",
    "rating_pred = pd.DataFrame(index = rating_test.index, columns = ['dummy']).join(rating_pred_mfnn.pivot(index='movieId', columns='userId', values='rating'), how = 'left').drop('dummy', axis = 1)\n",
    "\n",
    "\n",
    "#recall\n",
    "user_recalls, avg_recall = recall_over_user(rating_test, rating_pred, k_items=k_items)\n",
    "print(f\"Mean Recall: {avg_recall:.3f}\")\n",
    "user_grouped = group(train_df, 'user')\n",
    "recall_dict = recall_over_user_group(rating_test, rating_pred, user_grouped=user_grouped, k_items=k_items)\n",
    "\n",
    "#coverage\n",
    "user_cov = cal_coverage(rating_test, rating_pred, k_items=k_items, threshold=1, coverage_type='user')\n",
    "movie_cov = cal_coverage(rating_test, rating_pred, k_items=k_items, threshold=1, coverage_type = 'item')\n",
    "catalog_cov = cal_coverage(rating_test, rating_pred, k_items=k_items, threshold=1, coverage_type = 'catalog')\n",
    "\n",
    "#ndcg\n",
    "ndcg_score = ndcg(rating_test, rating_pred)\n",
    "print()\n",
    "print(f\"NDCG: {ndcg_score:.3f}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<b>Result Interpretation</b>: \n",
    "\n",
    "As the result shows, adding in side information of movie tag relavance fails to boost performances in all metrics. And it performs very poor for the most active user groups, perhaps it has also introduced noises and could not fit well into this pipeline."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 5.2 ANN-NN Model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Algorithms deployed:\n",
    "\n",
    "- Preranking stage: ANN (kd-tree based)\n",
    "- Reranking stage: NN"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 5.2.1 ANN-NN Model (MF embedding)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training begins.......\n",
      "Begin to retrieve latent features from MF ......\n",
      "Begin to retrieve top 20 pairs with ann......\n",
      "top 20 pairs are ready\n",
      "creating training set for neural network......\n",
      "shape of features array in train set: (952005, 40)\n",
      "shape of target array in train set: (952005,)\n",
      "Prediction begins.......\n",
      "creating testing set for neural network.....\n",
      "(399930, 4)\n",
      "shape of features array in test set: (399930, 40)\n",
      "shape of target array in test set: (399930,)\n",
      "train_time:  400.2902810573578\n",
      "test_time:  40.054936170578\n"
     ]
    }
   ],
   "source": [
    "#model initializing\n",
    "\n",
    "k_items=10\n",
    "#training\n",
    "start = time.time()  \n",
    "model_ann_nn = ann_nn(max_iter=2, regparam=0.1, prerank_k=20, leaf_size=30, rank=20, \n",
    "                      batch_size=256, epochs=50, k_items=k_items, verbose = 0, use_element_embed=False)\n",
    "model_ann_nn.fit(train_df, df_tag)\n",
    "end = time.time()  \n",
    "train_time = end - start\n",
    "\n",
    "#testing\n",
    "start = time.time()  \n",
    "df_pred =  model_ann_nn.pred(test_df)\n",
    "end = time.time() \n",
    "test_time = end - start \n",
    "\n",
    "print(\"train_time: \", train_time)\n",
    "print(\"test_time: \", test_time)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_history(model_ann_nn.history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean Recall: 0.065\n",
      "Recall for each user group:  {'10-50': 0.0412, '50-100': 0.0439, '100-500': 0.0449, '500-1000': 0.0667}\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "User coverage:  the fraction of users for which at least 1 items can be recommended well is 0.354\n",
      "Item coverage: the fraction of items that can be recommended to at least 1 users well is 0.311\n",
      "Catalog coverage: the fraction of items that are in the top-10 for at least 1 user is 0.311\n",
      "\n",
      "NDCG: 0.339\n"
     ]
    }
   ],
   "source": [
    "# Evaluation\n",
    "\n",
    "k_items = 10\n",
    "rating_pred = pd.DataFrame(index = rating_test.index, columns = ['dummy']).join(df_pred.pivot(index='movieId', columns='userId', values='rating'), how = 'left').drop('dummy', axis = 1)\n",
    "\n",
    "#recall\n",
    "user_recalls, avg_recall = recall_over_user(rating_test, rating_pred, k_items=k_items)\n",
    "print(f\"Mean Recall: {avg_recall:.3f}\")\n",
    "user_grouped = group(train_df, 'user')\n",
    "recall_dict = recall_over_user_group(rating_test, rating_pred, user_grouped=user_grouped, k_items=k_items)\n",
    "\n",
    "#coverage\n",
    "user_cov = cal_coverage(rating_test, rating_pred, k_items=k_items, threshold=1, coverage_type='user')\n",
    "movie_cov = cal_coverage(rating_test, rating_pred, k_items=k_items, threshold=1, coverage_type = 'item')\n",
    "catalog_cov = cal_coverage(rating_test, rating_pred, k_items=k_items, threshold=1, coverage_type = 'catalog')\n",
    "\n",
    "#ndcg\n",
    "ndcg_score = ndcg(rating_test, rating_pred)\n",
    "print()\n",
    "print(f\"NDCG: {ndcg_score:.3f}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<b>Result Interpretation</b>: \n",
    "\n",
    "As the result shows, using ANN-NN hybrid model performs the best in all key metrics so far: `recall@10 = 6.5%`, `user coverage = 35.4%`, `NDCG = 0.339`.\n",
    "\n",
    "It also predicts better as ratings of movie increase for each user group. \n",
    "\n",
    "\n",
    "Although movie coverage is a bit lower than the single ANN model, it performs quite well in mean recall@10 for all user groups (no less than 0.04).\n",
    " \n",
    " \n",
    "The complexity introduced and more time taken to train could indeed result in improvement in mean recall@10, which is worth the efforts."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 5.2.1 ANN-NN Model (MF embedding + movie tag embedding)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training begins.......\n",
      "Begin to retrieve latent features from MF ......\n",
      "Begin to retrieve top 20 pairs with ann......\n",
      "top 20 pairs are ready\n",
      "creating training set for neural network......\n",
      "shape of features array in train set: (952005, 1168)\n",
      "shape of target array in train set: (952005,)\n",
      "Prediction begins.......\n",
      "creating testing set for neural network.....\n",
      "(399876, 4)\n",
      "shape of features array in test set: (399876, 1168)\n",
      "shape of target array in test set: (399876,)\n",
      "train_time:  1185.3029670715332\n",
      "test_time:  78.87011003494263\n"
     ]
    }
   ],
   "source": [
    "#model initializing\n",
    "\n",
    "k_items=10\n",
    "#training\n",
    "start = time.time()  \n",
    "model_ann_nn = ann_nn(max_iter=2, regparam=0.1, prerank_k=20, leaf_size=30, rank=20, \n",
    "                      batch_size=256, epochs=32, k_items=k_items, verbose = 0, use_element_embed=True)\n",
    "model_ann_nn.fit(train_df, df_tag)\n",
    "end = time.time()  \n",
    "train_time = end - start\n",
    "\n",
    "#testing\n",
    "start = time.time()  \n",
    "df_pred =  model_ann_nn.pred(test_df)\n",
    "end = time.time() \n",
    "test_time = end - start \n",
    "\n",
    "print(\"train_time: \", train_time)\n",
    "print(\"test_time: \", test_time)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_history(model_ann_nn.history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean Recall: 0.082\n",
      "Recall for each user group:  {'10-50': 0.056, '50-100': 0.0583, '100-500': 0.0595, '500-1000': 0.0667}\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "User coverage:  the fraction of users for which at least 1 items can be recommended well is 0.427\n",
      "Item coverage: the fraction of items that can be recommended to at least 1 users well is 0.350\n",
      "Catalog coverage: the fraction of items that are in the top-10 for at least 1 user is 0.350\n",
      "\n",
      "NDCG: 0.356\n"
     ]
    }
   ],
   "source": [
    "# Evaluation\n",
    "\n",
    "k_items = 10\n",
    "rating_pred = pd.DataFrame(index = rating_test.index, columns = ['dummy']).join(df_pred.pivot(index='movieId', columns='userId', values='rating'), how = 'left').drop('dummy', axis = 1)\n",
    "\n",
    "#recall\n",
    "user_recalls, avg_recall = recall_over_user(rating_test, rating_pred, k_items=k_items)\n",
    "print(f\"Mean Recall: {avg_recall:.3f}\")\n",
    "user_grouped = group(train_df, 'user')\n",
    "recall_dict = recall_over_user_group(rating_test, rating_pred, user_grouped=user_grouped, k_items=k_items)\n",
    "\n",
    "#coverage\n",
    "user_cov = cal_coverage(rating_test, rating_pred, k_items=k_items, threshold=1, coverage_type='user')\n",
    "movie_cov = cal_coverage(rating_test, rating_pred, k_items=k_items, threshold=1, coverage_type = 'item')\n",
    "catalog_cov = cal_coverage(rating_test, rating_pred, k_items=k_items, threshold=1, coverage_type = 'catalog')\n",
    "\n",
    "#ndcg\n",
    "ndcg_score = ndcg(rating_test, rating_pred)\n",
    "print()\n",
    "print(f\"NDCG: {ndcg_score:.3f}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<b>Result Interpretation</b>: \n",
    "\n",
    "Adding in side information of movie tag relavance DID boost performances in all key metrics. It achieves the best result so far:\n",
    "`recall@10 = 8.2%`, `user coverage=42.7%`, `NDCG=0.356`.\n",
    "\n",
    "And it achieves good recall score on all user groups (no less than 5%) \n",
    "\n",
    "This indicates that adding side information upon the simple algorithm like ANN, could indeed help improve model' ability to predict well in the re-ranking stage. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 5. Conclusion"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<b>Our Goal</b>\n",
    "\n",
    "- Our overall objective is to build a production-quality movie recommendation pipeline, in order to `provide personalizied TOP-10 list of movies specific to our users`.\n",
    "\n",
    "<b>What We Have Explored</b>\n",
    "\n",
    "- We have explored and tuned different single and hybrid models. For benchmarking, we built 2 baseline models - `Bias Model`, `Matrix Factorization Model(MF)`. \n",
    "\n",
    "- Built upon movie and user embedding generated by MF, we also explored another efficient algorithm - `KDTree-based Approximate Nearest Neighbor(ANN)`. \n",
    "\n",
    "- To empower our prediction of ranking with stronger algorithms, we built 2 hybrid models - `MF-NN model` and `ANN-NN model`. This combination of models allow an efficient model to find the approximate TOP K first, and then use a stronger model like Neural Network to finally choose TOP 10 out of it. \n",
    "\n",
    "- For the `Neural Network(NN)` model, we mainly fed into it the `user-movie embeddings` generated by MF model. And we also introduced side information like `tag relevance embedding` to further improve its potential to predict the ratings well.\n",
    "\n",
    "\n",
    "<b>Final Recommendation</b>\n",
    "\n",
    "- To decide whether to launch these models, we evaluated their performances on our key metrics - `recall@10`, `user coverage@1` and `scalability`. The results of each model explored are demonstrated in the table below.\n",
    "\n",
    "- `ANN-NN Model with MF & Tag embedding` performs the best in both recall@10 and user coverage@1. These two results mean that the recommendation system would recommend at least 1 movie well to `42.7%` percent of active users, and would recommend the true TOP 10 rated movies for `8.2%`of the time. The NDCG score also ranks the highest. And the training time and test time is within the acceptable range, which takes about 20 minutes to train and 1 minute to predict. This would be sufficient for a daily update. \n",
    "\n",
    "- With this result, we would recommend the technical manager to move forward with the `ANN-NN Model with MF & Tag embedding` for AB testing and launching.\n",
    "\n",
    "<b>Watchouts and Future Exploration</b>\n",
    "\n",
    "- We are also aware that as number of target users or the update frequency requirement increases, we may need to test using other more efficient models like `Scalable Nearest Neighbor(ScaNN)` to scale with large datasets. \n",
    "\n",
    "- Other things we could explore include using other sophisticated models like FM structures and different features like movie genres and movie description text.\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n",
    "\n",
    "![jupyter](./result1.png)\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}