RecommSystems_R5_Project1.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Problem Statement\n",
    "\n",
    "Build your own recommendation system for products on an e-commerce website like Amazon.com.\n",
    "Dataset Link - https://drive.google.com/file/d/1ClBptsK3V5KgKXtK2GSRzFNAW7GnTPDW/view?usp=sharing\n",
    "\n",
    " \n",
    "\n",
    "Dataset columns - first three columns are userId, productId, and ratings and the fourth column is timestamp. You can discard the timestamp column as in this case you may not need to use it.\n",
    "\n",
    "Source - Amazon Reviews data (http://jmcauley.ucsd.edu/data/amazon/)  The repository has several datasets. For this case study, we are using the Electronics dataset."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Steps\n",
    "\n",
    "Please do the analysis based on steps (1 to 8) as given below:\n",
    "\n",
    "- Read and explore the given dataset. (Rename column/add headers, plot histograms, find data characteristics)\n",
    "- Take a subset of the dataset to make it less sparse/ denser. (For example, keep the users only who has given 50 or more number of ratings)\n",
    "- Split the data randomly into train and test dataset. (For example, split it in 70/30 ratio)\n",
    "- Build Popularity Recommender model.\n",
    "- Build Collaborative Filtering model.\n",
    "- Evaluate both the models. (Once the model is trained on the training data, it can be used to compute the error (like RMSE) on predictions made on the test data.) You can also use a different method to evaluate the models.\n",
    "- Get top - K (K = 5) recommendations. Since our goal is to recommend new products to each user based on his/her habits, we will recommend 5 new products.\n",
    "- Summarise your insights."
   ]
  },
  {
   "attachments": {
    "image.png": {
     "image/png": "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"
    }
   },
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![image.png](attachment:image.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Please Note**\n",
    "\n",
    "If you are facing any memory issue while working on this project, create a small subset (Let’s say 10% of data) and work on it.\n",
    "If you are stuck at the model evaluation part of this project.\n",
    "Please refer to below links - \n",
    "\n",
    "https://surprise.readthedocs.io/en/stable/accuracy.html (Links to an external site.)Links to an external site.\n",
    "http://surpriselib.com/ (Links to an external site.)Links to an external site. - Getting started, example"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Import required libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import seaborn as sns\n",
    "from collections import defaultdict\n",
    "from scipy.sparse.linalg import svds\n",
    "\n",
    "from surprise import SVD, KNNWithMeans\n",
    "from surprise import Dataset, Reader\n",
    "from surprise import accuracy\n",
    "from surprise.model_selection import train_test_split\n",
    "from surprise.model_selection import GridSearchCV\n",
    "\n",
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. Read and explore the given dataset. (Rename column/add headers, plot histograms, find data characteristics)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "#Loading data\n",
    "ratings = pd.read_csv(\"Data/ratings_Electronics.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "ratings.columns = ['userId', 'productId', 'ratings', 'timestamp']"
   ]
  },
  {
   "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>userId</th>\n",
       "      <th>productId</th>\n",
       "      <th>ratings</th>\n",
       "      <th>timestamp</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>A2CX7LUOHB2NDG</td>\n",
       "      <td>0321732944</td>\n",
       "      <td>5.0</td>\n",
       "      <td>1341100800</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>A2NWSAGRHCP8N5</td>\n",
       "      <td>0439886341</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1367193600</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>A2WNBOD3WNDNKT</td>\n",
       "      <td>0439886341</td>\n",
       "      <td>3.0</td>\n",
       "      <td>1374451200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>A1GI0U4ZRJA8WN</td>\n",
       "      <td>0439886341</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1334707200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>A1QGNMC6O1VW39</td>\n",
       "      <td>0511189877</td>\n",
       "      <td>5.0</td>\n",
       "      <td>1397433600</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           userId   productId  ratings   timestamp\n",
       "0  A2CX7LUOHB2NDG  0321732944      5.0  1341100800\n",
       "1  A2NWSAGRHCP8N5  0439886341      1.0  1367193600\n",
       "2  A2WNBOD3WNDNKT  0439886341      3.0  1374451200\n",
       "3  A1GI0U4ZRJA8WN  0439886341      1.0  1334707200\n",
       "4  A1QGNMC6O1VW39  0511189877      5.0  1397433600"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ratings.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "ratings.drop('timestamp', axis=1, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "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>productId</th>\n",
       "      <th>ratings</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>A2CX7LUOHB2NDG</td>\n",
       "      <td>0321732944</td>\n",
       "      <td>5.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>A2NWSAGRHCP8N5</td>\n",
       "      <td>0439886341</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>A2WNBOD3WNDNKT</td>\n",
       "      <td>0439886341</td>\n",
       "      <td>3.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>A1GI0U4ZRJA8WN</td>\n",
       "      <td>0439886341</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>A1QGNMC6O1VW39</td>\n",
       "      <td>0511189877</td>\n",
       "      <td>5.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           userId   productId  ratings\n",
       "0  A2CX7LUOHB2NDG  0321732944      5.0\n",
       "1  A2NWSAGRHCP8N5  0439886341      1.0\n",
       "2  A2WNBOD3WNDNKT  0439886341      3.0\n",
       "3  A1GI0U4ZRJA8WN  0439886341      1.0\n",
       "4  A1QGNMC6O1VW39  0511189877      5.0"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ratings.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "userId       7824481\n",
       "productId    7824481\n",
       "ratings      7824481\n",
       "dtype: int64"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ratings.count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "ratings.drop_duplicates(inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### We have taken a subset of the data as it takes large CPU/memory for the entire dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "ratings_small = ratings.sample(frac=0.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x17780a4bcc0>"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "sns.countplot(ratings_small.ratings)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. Take a subset of the dataset to make it less sparse/ denser. ( For example, keep the users only who has given 50 or more number of ratings)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "userGroups = ratings_small.groupby(by='userId').count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "user50 = userGroups.loc[userGroups['productId'] > 49].index"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "ratings_new = ratings_small.loc[ratings_small.userId.isin (user50)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(22264, 3)"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ratings_new.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x1778149dc18>"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "sns.countplot(ratings_new.ratings)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3. Split the data randomly into train and test dataset. (For example, split it in 70/30 ratio)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = Dataset.load_from_df(ratings_new, Reader())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "trainset, testset = train_test_split(data, test_size=.30)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4. Build Popularity Recommender model."
   ]
  },
  {
   "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>productId</th>\n",
       "      <th>ratings</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1602109</th>\n",
       "      <td>A20DZX38KRBIT8</td>\n",
       "      <td>B0012GQZZU</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7141376</th>\n",
       "      <td>A1V4VVBQBFXRHC</td>\n",
       "      <td>B00BOHNYTW</td>\n",
       "      <td>5.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6185788</th>\n",
       "      <td>A2AF0NOCM71J0B</td>\n",
       "      <td>B008DBI6Y0</td>\n",
       "      <td>4.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>407591</th>\n",
       "      <td>AT6CZDCP4TRGA</td>\n",
       "      <td>B0001U6RNI</td>\n",
       "      <td>5.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4178566</th>\n",
       "      <td>A3IY316DRNF5F2</td>\n",
       "      <td>B004IAPSAA</td>\n",
       "      <td>5.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 userId   productId  ratings\n",
       "1602109  A20DZX38KRBIT8  B0012GQZZU      1.0\n",
       "7141376  A1V4VVBQBFXRHC  B00BOHNYTW      5.0\n",
       "6185788  A2AF0NOCM71J0B  B008DBI6Y0      4.0\n",
       "407591    AT6CZDCP4TRGA  B0001U6RNI      5.0\n",
       "4178566  A3IY316DRNF5F2  B004IAPSAA      5.0"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ratings_new.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Let's group the books and find out mean ratings for every book"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "book_ratings = pd.DataFrame(ratings_new.groupby(by='productId', axis=0).mean())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Let's group the books and find out count of ratings for every book and add to book_ratings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "book_ratings['count'] = ratings_new.groupby(by='productId', axis=0).count()['ratings']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Sorting this table by mean ratings and by number of ratings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "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>ratings</th>\n",
       "      <th>count</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>productId</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>B003A7CDY6</th>\n",
       "      <td>5.0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>B003XKWLSS</th>\n",
       "      <td>5.0</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>B003XDU2XG</th>\n",
       "      <td>5.0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>B003XDU2Y0</th>\n",
       "      <td>5.0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>B003XESDFO</th>\n",
       "      <td>5.0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            ratings  count\n",
       "productId                 \n",
       "B003A7CDY6      5.0      1\n",
       "B003XKWLSS      5.0      3\n",
       "B003XDU2XG      5.0      1\n",
       "B003XDU2Y0      5.0      1\n",
       "B003XESDFO      5.0      1"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "book_ratings.sort_values(by='ratings', ascending=False).head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "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>ratings</th>\n",
       "      <th>count</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>productId</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>B0088CJT4U</th>\n",
       "      <td>4.052632</td>\n",
       "      <td>38</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>B002R5AM7C</th>\n",
       "      <td>4.083333</td>\n",
       "      <td>36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>B00829TIEK</th>\n",
       "      <td>4.678571</td>\n",
       "      <td>28</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>B008DWCRQW</th>\n",
       "      <td>4.111111</td>\n",
       "      <td>27</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>B004CLYEDC</th>\n",
       "      <td>4.703704</td>\n",
       "      <td>27</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "             ratings  count\n",
       "productId                  \n",
       "B0088CJT4U  4.052632     38\n",
       "B002R5AM7C  4.083333     36\n",
       "B00829TIEK  4.678571     28\n",
       "B008DWCRQW  4.111111     27\n",
       "B004CLYEDC  4.703704     27"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "book_ratings.sort_values(by='count', ascending=False).head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### We can suggest Recommendations from above table either based on ratings or based on number of ratings given or a combination of both"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 5. Build Collaborative Filtering model."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Building and training the model with these best parameters"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "There are two ways of Collaborative Filtering: Either user-user or item-item.\n",
    "\n",
    "Here, we will first do the user-user based Collborative filtering and build the model with optimal parameters.\n",
    "\n",
    "Then, we will do the item-item based Collaborative filtering and build the model with optimal parameters."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### A. User-User Collaborative Filtering:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "user_based_param_grid = {'k': [20, 40],\n",
    "                         'sim_options': {'name': ['pearson_baseline', 'cosine'],\n",
    "                                         'user_based': [True]}}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "gs = GridSearchCV(KNNWithMeans, user_based_param_grid, measures=['rmse'], cv=3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Estimating biases using als...\n",
      "Computing the pearson_baseline similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Estimating biases using als...\n",
      "Computing the pearson_baseline similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Estimating biases using als...\n",
      "Computing the pearson_baseline similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Computing the cosine similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Computing the cosine similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Computing the cosine similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Estimating biases using als...\n",
      "Computing the pearson_baseline similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Estimating biases using als...\n",
      "Computing the pearson_baseline similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Estimating biases using als...\n",
      "Computing the pearson_baseline similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Computing the cosine similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Computing the cosine similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Computing the cosine similarity matrix...\n",
      "Done computing similarity matrix.\n"
     ]
    }
   ],
   "source": [
    "gs.fit(data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "({'rmse': 1.0089574127784655},\n",
       " {'rmse': {'k': 20,\n",
       "   'sim_options': {'name': 'pearson_baseline', 'user_based': True}}})"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gs.best_score, gs.best_params"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "user_based_model = KNNWithMeans(k=gs.best_params['rmse']['k'],\n",
    "                                sim_options=gs.best_params['rmse']['sim_options'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Estimating biases using als...\n",
      "Computing the pearson_baseline similarity matrix...\n",
      "Done computing similarity matrix.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<surprise.prediction_algorithms.knns.KNNWithMeans at 0x17780a3bcf8>"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "user_based_model.fit(trainset)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### B. Item-Item Collaborative Filtering Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "item_based_param_grid = {'k': [20, 40],\n",
    "                         'sim_options': {'name': ['pearson_baseline', 'cosine'],\n",
    "                                         'user_based': [False]}}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "gs = GridSearchCV(KNNWithMeans, item_based_param_grid, measures=['rmse'], cv=3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Estimating biases using als...\n",
      "Computing the pearson_baseline similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Estimating biases using als...\n",
      "Computing the pearson_baseline similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Estimating biases using als...\n",
      "Computing the pearson_baseline similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Computing the cosine similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Computing the cosine similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Computing the cosine similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Estimating biases using als...\n",
      "Computing the pearson_baseline similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Estimating biases using als...\n",
      "Computing the pearson_baseline similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Estimating biases using als...\n",
      "Computing the pearson_baseline similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Computing the cosine similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Computing the cosine similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Computing the cosine similarity matrix...\n",
      "Done computing similarity matrix.\n"
     ]
    }
   ],
   "source": [
    "gs.fit(data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "({'rmse': 1.0725635184138864},\n",
       " {'rmse': {'k': 20,\n",
       "   'sim_options': {'name': 'pearson_baseline', 'user_based': False}}})"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gs.best_score, gs.best_params"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "item_based_model = KNNWithMeans(k=gs.best_params['rmse']['k'],\n",
    "                                sim_options=gs.best_params['rmse']['sim_options'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Estimating biases using als...\n",
      "Computing the pearson_baseline similarity matrix...\n",
      "Done computing similarity matrix.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<surprise.prediction_algorithms.knns.KNNWithMeans at 0x1778152e4e0>"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "item_based_model.fit(trainset)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 6. Evaluate both the models. (Once the model is trained on the training data, it can be used to compute the error (like RMSE) on predictions made on the test data.) You can also use a different method to evaluate the models."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "user_pred = user_based_model.test(testset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy of the user-user Collaborative Filtering Model:\n",
      "RMSE: 1.0145\n",
      "MAE:  0.7729\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(1.014501217774407, 0.7728959137330035)"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print('Accuracy of the user-user Collaborative Filtering Model:')\n",
    "accuracy.rmse(user_pred), accuracy.mae(user_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "item_pred = item_based_model.test(testset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy of the item-item Collaborative Filtering Model:\n",
      "RMSE: 1.0785\n",
      "MAE:  0.7841\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(1.078531981694536, 0.7840748185382526)"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print('Accuracy of the item-item Collaborative Filtering Model:')\n",
    "accuracy.rmse(item_pred), accuracy.mae(item_pred)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Evaluation Summary:\n",
    "\n",
    "- Poulation model cannot be quantitatively assessed unlike the Collaborative filtering model.\n",
    "- Accuracy of the Collaborative filtering model can be assessed by RMSE (or MAE) value.\n",
    "- RMSE (or MAE) value should be less for any model.\n",
    "- Here, RMSE (or MAE) value is lower for user-user model than item-item model.\n",
    "- Moreover, we observed that item-item based model takes more CPU/memory than user-user based. May be due to large number of items than users.\n",
    "- So, user-user model is more accurate and efficient than the item-item model.\n",
    "- RMSE of both collaborative filtering model is almost 1; seems, the model needs further tuning "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 7. Get top - K (K = 5) recommendations. Since our goal is to recommend new products to each user based on his/her habits, we will recommend 5 new products."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Function for getting top5 prediections from the model.test() prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_top_5(predictions):\n",
    "\n",
    "    reco = []\n",
    "    # First map the predictions to each user.\n",
    "    top_n = defaultdict(list)\n",
    "    for uid, iid, true_r, est, _ in predictions:\n",
    "        top_n[uid].append((iid, est))\n",
    "\n",
    "    # Then sort the predictions for each user and retrieve the k highest ones.\n",
    "    for uid, user_ratings in top_n.items():\n",
    "        user_ratings.sort(key=lambda x: x[1], reverse=True)\n",
    "        \n",
    "        top_5_per_user = []\n",
    "        top_5_per_user.append(uid)\n",
    "        for ii in range(5):\n",
    "            top_5_per_user.append(user_ratings[ii][0])\n",
    "            #print(user_ratings[ii][1])\n",
    "        reco.append(top_5_per_user)\n",
    "\n",
    "    reco_df = pd.DataFrame(reco, columns=['userId', 'Reco1', 'Reco2', 'Reco3', 'Reco4', 'Reco5'])\n",
    "    return reco_df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Building anti_testset for user-item combinations which are not yet rated. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "anti_test = trainset.build_anti_testset()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### A. Predictions from user-user based Model:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "pred_anti_test = user_based_model.test(anti_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "top5_user_based_df = get_top_5(pred_anti_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "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>Reco1</th>\n",
       "      <th>Reco2</th>\n",
       "      <th>Reco3</th>\n",
       "      <th>Reco4</th>\n",
       "      <th>Reco5</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>A3LGT6UZL99IW1</td>\n",
       "      <td>B0080EBMM8</td>\n",
       "      <td>B000BI2162</td>\n",
       "      <td>B005QX7KYU</td>\n",
       "      <td>B004N625BY</td>\n",
       "      <td>B0002CPBWI</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>A2ARDVMUGDFED6</td>\n",
       "      <td>B004NY9UV4</td>\n",
       "      <td>B008X9ZBVI</td>\n",
       "      <td>B006JD3F0Y</td>\n",
       "      <td>B0009SV24G</td>\n",
       "      <td>B000N99BBC</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>A250AXLRBVYKB4</td>\n",
       "      <td>B00CI3BIWQ</td>\n",
       "      <td>B006LW0W5Y</td>\n",
       "      <td>B002IO2UM2</td>\n",
       "      <td>B003GDFJUO</td>\n",
       "      <td>B000095SB4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>A2BLFCOPSMBOZ9</td>\n",
       "      <td>B00JWXBM62</td>\n",
       "      <td>B001E1Y5O6</td>\n",
       "      <td>B003UH0Z9Q</td>\n",
       "      <td>B00FP3OX3C</td>\n",
       "      <td>B002MAPSQM</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>A2R6RA8FRBS608</td>\n",
       "      <td>B00007E7QS</td>\n",
       "      <td>B002BH4MEA</td>\n",
       "      <td>B00BQ4F9ZA</td>\n",
       "      <td>B000TFIWQK</td>\n",
       "      <td>B000VE4PNS</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           userId       Reco1       Reco2       Reco3       Reco4       Reco5\n",
       "0  A3LGT6UZL99IW1  B0080EBMM8  B000BI2162  B005QX7KYU  B004N625BY  B0002CPBWI\n",
       "1  A2ARDVMUGDFED6  B004NY9UV4  B008X9ZBVI  B006JD3F0Y  B0009SV24G  B000N99BBC\n",
       "2  A250AXLRBVYKB4  B00CI3BIWQ  B006LW0W5Y  B002IO2UM2  B003GDFJUO  B000095SB4\n",
       "3  A2BLFCOPSMBOZ9  B00JWXBM62  B001E1Y5O6  B003UH0Z9Q  B00FP3OX3C  B002MAPSQM\n",
       "4  A2R6RA8FRBS608  B00007E7QS  B002BH4MEA  B00BQ4F9ZA  B000TFIWQK  B000VE4PNS"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "top5_user_based_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### B. Predictions from item-item based Model:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "pred_anti_test = item_based_model.test(anti_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "top5_item_based_df = get_top_5(pred_anti_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "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>Reco1</th>\n",
       "      <th>Reco2</th>\n",
       "      <th>Reco3</th>\n",
       "      <th>Reco4</th>\n",
       "      <th>Reco5</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>A3LGT6UZL99IW1</td>\n",
       "      <td>B002BH4MEA</td>\n",
       "      <td>B000TFIWQK</td>\n",
       "      <td>B000VE4PNS</td>\n",
       "      <td>B0001FV3C8</td>\n",
       "      <td>B001RNNXQM</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>A2ARDVMUGDFED6</td>\n",
       "      <td>B000TFIWQK</td>\n",
       "      <td>B000VE4PNS</td>\n",
       "      <td>B0001FV3C8</td>\n",
       "      <td>B001RNNXQM</td>\n",
       "      <td>B000GA7048</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>A250AXLRBVYKB4</td>\n",
       "      <td>B002BH4MEA</td>\n",
       "      <td>B000TFIWQK</td>\n",
       "      <td>B000VE4PNS</td>\n",
       "      <td>B0001FV3C8</td>\n",
       "      <td>B001RNNXQM</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>A2BLFCOPSMBOZ9</td>\n",
       "      <td>B002BH4MEA</td>\n",
       "      <td>B000VE4PNS</td>\n",
       "      <td>B0001FV3C8</td>\n",
       "      <td>B001RNNXQM</td>\n",
       "      <td>B000GA7048</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>A2R6RA8FRBS608</td>\n",
       "      <td>B002BH4MEA</td>\n",
       "      <td>B000TFIWQK</td>\n",
       "      <td>B000VE4PNS</td>\n",
       "      <td>B0001FV3C8</td>\n",
       "      <td>B001RNNXQM</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           userId       Reco1       Reco2       Reco3       Reco4       Reco5\n",
       "0  A3LGT6UZL99IW1  B002BH4MEA  B000TFIWQK  B000VE4PNS  B0001FV3C8  B001RNNXQM\n",
       "1  A2ARDVMUGDFED6  B000TFIWQK  B000VE4PNS  B0001FV3C8  B001RNNXQM  B000GA7048\n",
       "2  A250AXLRBVYKB4  B002BH4MEA  B000TFIWQK  B000VE4PNS  B0001FV3C8  B001RNNXQM\n",
       "3  A2BLFCOPSMBOZ9  B002BH4MEA  B000VE4PNS  B0001FV3C8  B001RNNXQM  B000GA7048\n",
       "4  A2R6RA8FRBS608  B002BH4MEA  B000TFIWQK  B000VE4PNS  B0001FV3C8  B001RNNXQM"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "top5_item_based_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### C. Predictions from user-item based model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For the user-item model, we will use matrix factorization technique e.g. SVD.\n",
    "- We will create a user-product matrix\n",
    "- Apply SVD to find probability of every user-product combination\n",
    "- Remove products already rated by the user\n",
    "- List the remaining products in decreasing probablity order\n",
    "- Get the top 5 recommendations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "user_product = pd.pivot(ratings_new, index='userId', columns='productId', values='ratings').fillna(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th>productId</th>\n",
       "      <th>0970407998</th>\n",
       "      <th>1400501466</th>\n",
       "      <th>1400599997</th>\n",
       "      <th>7562434166</th>\n",
       "      <th>787988002X</th>\n",
       "      <th>9043413585</th>\n",
       "      <th>9800359788</th>\n",
       "      <th>9981719005</th>\n",
       "      <th>B000001OMN</th>\n",
       "      <th>B000001ON0</th>\n",
       "      <th>...</th>\n",
       "      <th>B00L0YFJHY</th>\n",
       "      <th>B00L1I727Y</th>\n",
       "      <th>B00L21HC7A</th>\n",
       "      <th>B00L2442H0</th>\n",
       "      <th>B00L26YDA4</th>\n",
       "      <th>B00L3YHF6O</th>\n",
       "      <th>B00L403O94</th>\n",
       "      <th>B00LGQ6HL8</th>\n",
       "      <th>B00LI4ZZO8</th>\n",
       "      <th>B00LKG1MC8</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",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>A100UD67AHFODS</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>A100WO06OQR8BQ</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>A10PEXB6XAQ5XF</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>A10Y058K7B96C6</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>A10ZFE6YE0UHW8</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 13955 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "productId       0970407998  1400501466  1400599997  7562434166  787988002X  \\\n",
       "userId                                                                       \n",
       "A100UD67AHFODS         0.0         0.0         0.0         0.0         0.0   \n",
       "A100WO06OQR8BQ         0.0         0.0         0.0         0.0         0.0   \n",
       "A10PEXB6XAQ5XF         0.0         0.0         0.0         0.0         0.0   \n",
       "A10Y058K7B96C6         0.0         0.0         0.0         0.0         0.0   \n",
       "A10ZFE6YE0UHW8         0.0         0.0         0.0         0.0         0.0   \n",
       "\n",
       "productId       9043413585  9800359788  9981719005  B000001OMN  B000001ON0  \\\n",
       "userId                                                                       \n",
       "A100UD67AHFODS         0.0         0.0         0.0         0.0         0.0   \n",
       "A100WO06OQR8BQ         0.0         0.0         0.0         0.0         0.0   \n",
       "A10PEXB6XAQ5XF         0.0         0.0         0.0         0.0         0.0   \n",
       "A10Y058K7B96C6         0.0         0.0         0.0         0.0         0.0   \n",
       "A10ZFE6YE0UHW8         0.0         0.0         0.0         0.0         0.0   \n",
       "\n",
       "productId       ...  B00L0YFJHY  B00L1I727Y  B00L21HC7A  B00L2442H0  \\\n",
       "userId          ...                                                   \n",
       "A100UD67AHFODS  ...         0.0         0.0         0.0         0.0   \n",
       "A100WO06OQR8BQ  ...         0.0         0.0         0.0         0.0   \n",
       "A10PEXB6XAQ5XF  ...         0.0         0.0         0.0         0.0   \n",
       "A10Y058K7B96C6  ...         0.0         0.0         0.0         0.0   \n",
       "A10ZFE6YE0UHW8  ...         0.0         0.0         0.0         0.0   \n",
       "\n",
       "productId       B00L26YDA4  B00L3YHF6O  B00L403O94  B00LGQ6HL8  B00LI4ZZO8  \\\n",
       "userId                                                                       \n",
       "A100UD67AHFODS         0.0         0.0         0.0         0.0         0.0   \n",
       "A100WO06OQR8BQ         0.0         0.0         0.0         0.0         0.0   \n",
       "A10PEXB6XAQ5XF         0.0         0.0         0.0         0.0         0.0   \n",
       "A10Y058K7B96C6         0.0         0.0         0.0         0.0         0.0   \n",
       "A10ZFE6YE0UHW8         0.0         0.0         0.0         0.0         0.0   \n",
       "\n",
       "productId       B00LKG1MC8  \n",
       "userId                      \n",
       "A100UD67AHFODS         0.0  \n",
       "A100WO06OQR8BQ         0.0  \n",
       "A10PEXB6XAQ5XF         0.0  \n",
       "A10Y058K7B96C6         0.0  \n",
       "A10ZFE6YE0UHW8         0.0  \n",
       "\n",
       "[5 rows x 13955 columns]"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "user_product.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "U, sigma, VT = svds(user_product, k=50)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "sigma = np.diag(sigma)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [],
   "source": [
    "users_item_prob = np.dot(np.dot(U, sigma), VT)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "user_item_prob_df = pd.DataFrame(users_item_prob, columns=user_product.columns, index=user_product.index)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "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>productId</th>\n",
       "      <th>0970407998</th>\n",
       "      <th>1400501466</th>\n",
       "      <th>1400599997</th>\n",
       "      <th>7562434166</th>\n",
       "      <th>787988002X</th>\n",
       "      <th>9043413585</th>\n",
       "      <th>9800359788</th>\n",
       "      <th>9981719005</th>\n",
       "      <th>B000001OMN</th>\n",
       "      <th>B000001ON0</th>\n",
       "      <th>...</th>\n",
       "      <th>B00L0YFJHY</th>\n",
       "      <th>B00L1I727Y</th>\n",
       "      <th>B00L21HC7A</th>\n",
       "      <th>B00L2442H0</th>\n",
       "      <th>B00L26YDA4</th>\n",
       "      <th>B00L3YHF6O</th>\n",
       "      <th>B00L403O94</th>\n",
       "      <th>B00LGQ6HL8</th>\n",
       "      <th>B00LI4ZZO8</th>\n",
       "      <th>B00LKG1MC8</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",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>A100UD67AHFODS</th>\n",
       "      <td>0.000489</td>\n",
       "      <td>0.000430</td>\n",
       "      <td>-0.000264</td>\n",
       "      <td>0.001023</td>\n",
       "      <td>0.001553</td>\n",
       "      <td>-0.004923</td>\n",
       "      <td>0.057848</td>\n",
       "      <td>0.052781</td>\n",
       "      <td>-0.000181</td>\n",
       "      <td>0.111762</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.015198</td>\n",
       "      <td>0.026217</td>\n",
       "      <td>-0.023921</td>\n",
       "      <td>0.080002</td>\n",
       "      <td>0.054941</td>\n",
       "      <td>0.114335</td>\n",
       "      <td>-0.015208</td>\n",
       "      <td>0.013913</td>\n",
       "      <td>0.110153</td>\n",
       "      <td>-0.002677</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>A100WO06OQR8BQ</th>\n",
       "      <td>-0.004651</td>\n",
       "      <td>0.000426</td>\n",
       "      <td>-0.000145</td>\n",
       "      <td>-0.001642</td>\n",
       "      <td>0.042545</td>\n",
       "      <td>-0.024539</td>\n",
       "      <td>0.020221</td>\n",
       "      <td>-0.014188</td>\n",
       "      <td>0.000030</td>\n",
       "      <td>0.096839</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.016634</td>\n",
       "      <td>-0.017257</td>\n",
       "      <td>0.155870</td>\n",
       "      <td>-0.051489</td>\n",
       "      <td>0.004622</td>\n",
       "      <td>-0.154945</td>\n",
       "      <td>-0.004308</td>\n",
       "      <td>-0.231662</td>\n",
       "      <td>-0.073737</td>\n",
       "      <td>-0.012222</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>A10PEXB6XAQ5XF</th>\n",
       "      <td>-0.001531</td>\n",
       "      <td>0.000146</td>\n",
       "      <td>0.000050</td>\n",
       "      <td>-0.000824</td>\n",
       "      <td>0.014591</td>\n",
       "      <td>0.006480</td>\n",
       "      <td>-0.042209</td>\n",
       "      <td>0.049096</td>\n",
       "      <td>-0.000163</td>\n",
       "      <td>-0.021289</td>\n",
       "      <td>...</td>\n",
       "      <td>0.031729</td>\n",
       "      <td>-0.032557</td>\n",
       "      <td>0.159527</td>\n",
       "      <td>-0.073013</td>\n",
       "      <td>0.096618</td>\n",
       "      <td>0.009082</td>\n",
       "      <td>0.019849</td>\n",
       "      <td>0.008179</td>\n",
       "      <td>0.066227</td>\n",
       "      <td>0.007661</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>A10Y058K7B96C6</th>\n",
       "      <td>-0.001058</td>\n",
       "      <td>-0.000639</td>\n",
       "      <td>0.000646</td>\n",
       "      <td>0.007134</td>\n",
       "      <td>0.018343</td>\n",
       "      <td>0.094758</td>\n",
       "      <td>-0.030164</td>\n",
       "      <td>-0.000643</td>\n",
       "      <td>0.000947</td>\n",
       "      <td>-0.088918</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.063953</td>\n",
       "      <td>-0.055860</td>\n",
       "      <td>0.157971</td>\n",
       "      <td>-0.119873</td>\n",
       "      <td>-0.110414</td>\n",
       "      <td>0.238329</td>\n",
       "      <td>0.015878</td>\n",
       "      <td>0.139932</td>\n",
       "      <td>0.107829</td>\n",
       "      <td>-0.016613</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>A10ZFE6YE0UHW8</th>\n",
       "      <td>-0.000940</td>\n",
       "      <td>-0.000251</td>\n",
       "      <td>-0.000068</td>\n",
       "      <td>0.000375</td>\n",
       "      <td>0.012610</td>\n",
       "      <td>0.002404</td>\n",
       "      <td>-0.016917</td>\n",
       "      <td>0.034766</td>\n",
       "      <td>-0.000199</td>\n",
       "      <td>0.100928</td>\n",
       "      <td>...</td>\n",
       "      <td>0.010524</td>\n",
       "      <td>-0.076285</td>\n",
       "      <td>0.459715</td>\n",
       "      <td>-0.114894</td>\n",
       "      <td>0.246568</td>\n",
       "      <td>-0.107591</td>\n",
       "      <td>-0.000528</td>\n",
       "      <td>-0.171905</td>\n",
       "      <td>-0.036989</td>\n",
       "      <td>0.007995</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 13955 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "productId       0970407998  1400501466  1400599997  7562434166  787988002X  \\\n",
       "userId                                                                       \n",
       "A100UD67AHFODS    0.000489    0.000430   -0.000264    0.001023    0.001553   \n",
       "A100WO06OQR8BQ   -0.004651    0.000426   -0.000145   -0.001642    0.042545   \n",
       "A10PEXB6XAQ5XF   -0.001531    0.000146    0.000050   -0.000824    0.014591   \n",
       "A10Y058K7B96C6   -0.001058   -0.000639    0.000646    0.007134    0.018343   \n",
       "A10ZFE6YE0UHW8   -0.000940   -0.000251   -0.000068    0.000375    0.012610   \n",
       "\n",
       "productId       9043413585  9800359788  9981719005  B000001OMN  B000001ON0  \\\n",
       "userId                                                                       \n",
       "A100UD67AHFODS   -0.004923    0.057848    0.052781   -0.000181    0.111762   \n",
       "A100WO06OQR8BQ   -0.024539    0.020221   -0.014188    0.000030    0.096839   \n",
       "A10PEXB6XAQ5XF    0.006480   -0.042209    0.049096   -0.000163   -0.021289   \n",
       "A10Y058K7B96C6    0.094758   -0.030164   -0.000643    0.000947   -0.088918   \n",
       "A10ZFE6YE0UHW8    0.002404   -0.016917    0.034766   -0.000199    0.100928   \n",
       "\n",
       "productId       ...  B00L0YFJHY  B00L1I727Y  B00L21HC7A  B00L2442H0  \\\n",
       "userId          ...                                                   \n",
       "A100UD67AHFODS  ...   -0.015198    0.026217   -0.023921    0.080002   \n",
       "A100WO06OQR8BQ  ...   -0.016634   -0.017257    0.155870   -0.051489   \n",
       "A10PEXB6XAQ5XF  ...    0.031729   -0.032557    0.159527   -0.073013   \n",
       "A10Y058K7B96C6  ...   -0.063953   -0.055860    0.157971   -0.119873   \n",
       "A10ZFE6YE0UHW8  ...    0.010524   -0.076285    0.459715   -0.114894   \n",
       "\n",
       "productId       B00L26YDA4  B00L3YHF6O  B00L403O94  B00LGQ6HL8  B00LI4ZZO8  \\\n",
       "userId                                                                       \n",
       "A100UD67AHFODS    0.054941    0.114335   -0.015208    0.013913    0.110153   \n",
       "A100WO06OQR8BQ    0.004622   -0.154945   -0.004308   -0.231662   -0.073737   \n",
       "A10PEXB6XAQ5XF    0.096618    0.009082    0.019849    0.008179    0.066227   \n",
       "A10Y058K7B96C6   -0.110414    0.238329    0.015878    0.139932    0.107829   \n",
       "A10ZFE6YE0UHW8    0.246568   -0.107591   -0.000528   -0.171905   -0.036989   \n",
       "\n",
       "productId       B00LKG1MC8  \n",
       "userId                      \n",
       "A100UD67AHFODS   -0.002677  \n",
       "A100WO06OQR8BQ   -0.012222  \n",
       "A10PEXB6XAQ5XF    0.007661  \n",
       "A10Y058K7B96C6   -0.016613  \n",
       "A10ZFE6YE0UHW8    0.007995  \n",
       "\n",
       "[5 rows x 13955 columns]"
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "user_item_prob_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "probabilities = []"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "for user in user_item_prob_df.index:\n",
    "    #user = 'A100WO06OQR8BQ'\n",
    "    top5_per_user = []\n",
    "    top5_per_user.append(user)\n",
    "\n",
    "    user_products = ratings_new.loc[ratings_new.userId == user]\n",
    "    #print(user_products.head(10))\n",
    "    user_predictions = user_item_prob_df.loc[user].sort_values(ascending=False)\n",
    "    #print(user_predictions.head(10))\n",
    "\n",
    "    reco = user_predictions.loc[~user_predictions.index.isin(user_products.productId)]\n",
    "    for ii in range(5):\n",
    "        top5_per_user.append(reco.index[ii])\n",
    "\n",
    "    probabilities.append(top5_per_user)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [],
   "source": [
    "user_item_pred_df = pd.DataFrame(probabilities, columns=['userId', 'Reco1', 'Reco2', 'Reco3', 'Reco4', 'Reco5'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "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>Reco1</th>\n",
       "      <th>Reco2</th>\n",
       "      <th>Reco3</th>\n",
       "      <th>Reco4</th>\n",
       "      <th>Reco5</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>A100UD67AHFODS</td>\n",
       "      <td>B003ES5ZUU</td>\n",
       "      <td>B00F0DD0I6</td>\n",
       "      <td>B002WE4HE2</td>\n",
       "      <td>B00DK2JQOQ</td>\n",
       "      <td>B002V88HFE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>A100WO06OQR8BQ</td>\n",
       "      <td>B004CLYEDC</td>\n",
       "      <td>B0088CJT4U</td>\n",
       "      <td>B0097CZHAU</td>\n",
       "      <td>B000N99BBC</td>\n",
       "      <td>B0097CXWUW</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>A10PEXB6XAQ5XF</td>\n",
       "      <td>B0088CJT4U</td>\n",
       "      <td>B004CLYEDC</td>\n",
       "      <td>B004CLYEFK</td>\n",
       "      <td>B002R5AM7C</td>\n",
       "      <td>B000N99BBC</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>A10Y058K7B96C6</td>\n",
       "      <td>B002R5AM7C</td>\n",
       "      <td>B009NHWVIA</td>\n",
       "      <td>B00BOHNYU6</td>\n",
       "      <td>B00829TIEK</td>\n",
       "      <td>B00834SJSK</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>A10ZFE6YE0UHW8</td>\n",
       "      <td>B002R5AM7C</td>\n",
       "      <td>B00KVNY2KA</td>\n",
       "      <td>B00834SJSK</td>\n",
       "      <td>B00D50UNRM</td>\n",
       "      <td>B00G500AE0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           userId       Reco1       Reco2       Reco3       Reco4       Reco5\n",
       "0  A100UD67AHFODS  B003ES5ZUU  B00F0DD0I6  B002WE4HE2  B00DK2JQOQ  B002V88HFE\n",
       "1  A100WO06OQR8BQ  B004CLYEDC  B0088CJT4U  B0097CZHAU  B000N99BBC  B0097CXWUW\n",
       "2  A10PEXB6XAQ5XF  B0088CJT4U  B004CLYEDC  B004CLYEFK  B002R5AM7C  B000N99BBC\n",
       "3  A10Y058K7B96C6  B002R5AM7C  B009NHWVIA  B00BOHNYU6  B00829TIEK  B00834SJSK\n",
       "4  A10ZFE6YE0UHW8  B002R5AM7C  B00KVNY2KA  B00834SJSK  B00D50UNRM  B00G500AE0"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "user_item_pred_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 8. Summarise your insights."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- *Popularity based model* gives recommendations only on the popularity ratings, either based on count of ratings or based on highest ratings.\n",
    "- *Collaborative* filtering can be dones by user-user and item-item based similarities. The model is created by training a given set of data. The same model is used to predict the ratings for products, which are not yet rated by the user.\n",
    "    - From the prediction of item-item model, we see that for many users, the item predictions are similar to each other. This is obvious as similarities of items are measured and items are arranged based on their neighbours. \n",
    "- *SVD Algorithm* is used to compute the probablity of ratings of all products, including the ones not yet rated by the user.\n",
    "- All models, user-user, item-item and user-item give different predictions for single user. This  is good as one single user should be given different recommendations based on different algorithms/choices."
   ]
  }
 ],
 "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": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}