recommendation_to_production.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 46,
   "id": "standing-behalf",
   "metadata": {},
   "outputs": [],
   "source": [
    "from copy import deepcopy\n",
    "\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "from matplotlib import pyplot as plt\n",
    "\n",
    "from lightfm.data import Dataset\n",
    "from lightfm import LightFM\n",
    "from lightfm.evaluation import precision_at_k, auc_score"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "passive-disco",
   "metadata": {},
   "source": [
    "# Load data"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "aggressive-transaction",
   "metadata": {},
   "source": [
    "Dataset is acquired from from Kaggle:\n",
    "https://www.kaggle.com/prajitdatta/movielens-100k-dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "electrical-transcription",
   "metadata": {},
   "source": [
    "## Interactions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "intended-thickness",
   "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>user_id</th>\n",
       "      <th>movie_id</th>\n",
       "      <th>rating</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>196</td>\n",
       "      <td>242</td>\n",
       "      <td>3</td>\n",
       "      <td>1997-12-04 15:55:49</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>186</td>\n",
       "      <td>302</td>\n",
       "      <td>3</td>\n",
       "      <td>1998-04-04 19:22:22</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>298</td>\n",
       "      <td>474</td>\n",
       "      <td>4</td>\n",
       "      <td>1998-01-07 14:20:06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>253</td>\n",
       "      <td>465</td>\n",
       "      <td>5</td>\n",
       "      <td>1998-04-03 18:34:27</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>305</td>\n",
       "      <td>451</td>\n",
       "      <td>3</td>\n",
       "      <td>1998-02-01 09:20:17</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99992</th>\n",
       "      <td>721</td>\n",
       "      <td>262</td>\n",
       "      <td>3</td>\n",
       "      <td>1997-10-18 01:14:45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99994</th>\n",
       "      <td>378</td>\n",
       "      <td>78</td>\n",
       "      <td>3</td>\n",
       "      <td>1997-11-20 20:16:16</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99995</th>\n",
       "      <td>880</td>\n",
       "      <td>476</td>\n",
       "      <td>3</td>\n",
       "      <td>1997-11-22 05:10:44</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99996</th>\n",
       "      <td>716</td>\n",
       "      <td>204</td>\n",
       "      <td>5</td>\n",
       "      <td>1997-11-17 19:39:03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99999</th>\n",
       "      <td>12</td>\n",
       "      <td>203</td>\n",
       "      <td>3</td>\n",
       "      <td>1997-11-19 17:13:03</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>82520 rows × 4 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       user_id  movie_id  rating                time\n",
       "0          196       242       3 1997-12-04 15:55:49\n",
       "1          186       302       3 1998-04-04 19:22:22\n",
       "5          298       474       4 1998-01-07 14:20:06\n",
       "7          253       465       5 1998-04-03 18:34:27\n",
       "8          305       451       3 1998-02-01 09:20:17\n",
       "...        ...       ...     ...                 ...\n",
       "99992      721       262       3 1997-10-18 01:14:45\n",
       "99994      378        78       3 1997-11-20 20:16:16\n",
       "99995      880       476       3 1997-11-22 05:10:44\n",
       "99996      716       204       5 1997-11-17 19:39:03\n",
       "99999       12       203       3 1997-11-19 17:13:03\n",
       "\n",
       "[82520 rows x 4 columns]"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "interactions = pd.read_csv(\"data/ml-100k/u.data\", sep=\"\\t\", header=None)\n",
    "interactions.columns = [\"user_id\", \"movie_id\", \"rating\", \"time\"]\n",
    "\n",
    "interactions[\"time\"] = pd.to_datetime(interactions[\"time\"],unit=\"s\")\n",
    "\n",
    "# Filter on ratings 3 or higher\n",
    "interactions = interactions[interactions[\"rating\"] >= 3]\n",
    "\n",
    "interactions"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "conditional-encounter",
   "metadata": {},
   "source": [
    "## User features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "appreciated-monitoring",
   "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>user_id</th>\n",
       "      <th>age</th>\n",
       "      <th>gender</th>\n",
       "      <th>occupation</th>\n",
       "      <th>zip_code</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>24</td>\n",
       "      <td>M</td>\n",
       "      <td>technician</td>\n",
       "      <td>85711</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>53</td>\n",
       "      <td>F</td>\n",
       "      <td>other</td>\n",
       "      <td>94043</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>23</td>\n",
       "      <td>M</td>\n",
       "      <td>writer</td>\n",
       "      <td>32067</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>24</td>\n",
       "      <td>M</td>\n",
       "      <td>technician</td>\n",
       "      <td>43537</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>33</td>\n",
       "      <td>F</td>\n",
       "      <td>other</td>\n",
       "      <td>15213</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>938</th>\n",
       "      <td>939</td>\n",
       "      <td>26</td>\n",
       "      <td>F</td>\n",
       "      <td>student</td>\n",
       "      <td>33319</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>939</th>\n",
       "      <td>940</td>\n",
       "      <td>32</td>\n",
       "      <td>M</td>\n",
       "      <td>administrator</td>\n",
       "      <td>02215</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>940</th>\n",
       "      <td>941</td>\n",
       "      <td>20</td>\n",
       "      <td>M</td>\n",
       "      <td>student</td>\n",
       "      <td>97229</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>941</th>\n",
       "      <td>942</td>\n",
       "      <td>48</td>\n",
       "      <td>F</td>\n",
       "      <td>librarian</td>\n",
       "      <td>78209</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>942</th>\n",
       "      <td>943</td>\n",
       "      <td>22</td>\n",
       "      <td>M</td>\n",
       "      <td>student</td>\n",
       "      <td>77841</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>943 rows × 5 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     user_id  age gender     occupation zip_code\n",
       "0          1   24      M     technician    85711\n",
       "1          2   53      F          other    94043\n",
       "2          3   23      M         writer    32067\n",
       "3          4   24      M     technician    43537\n",
       "4          5   33      F          other    15213\n",
       "..       ...  ...    ...            ...      ...\n",
       "938      939   26      F        student    33319\n",
       "939      940   32      M  administrator    02215\n",
       "940      941   20      M        student    97229\n",
       "941      942   48      F      librarian    78209\n",
       "942      943   22      M        student    77841\n",
       "\n",
       "[943 rows x 5 columns]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "user_features = pd.read_csv(\"data/ml-100k/u.user\", sep=\"|\", header=None)\n",
    "user_features.columns = [\"user_id\", \"age\", \"gender\", \"occupation\", \"zip_code\"]\n",
    "\n",
    "user_features"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "outside-muscle",
   "metadata": {},
   "source": [
    "## Item features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "detected-advisory",
   "metadata": {},
   "outputs": [],
   "source": [
    "item_features = pd.read_csv(\"data/ml-100k/u.item\", sep=\"|\", header=None)\n",
    "item_features.columns = [\n",
    "    \"movie_id\", \"movie_title\", \"release_date\", \"video_release_date\",\n",
    "    \"IMDb_URL\", \"unknown\", \"Action\", \"Adventure\", \"Animation\",\n",
    "    \"Children's\", \"Comedy\", \"Crime\", \"Documentary\", \"Drama\", \"Fantasy\",\n",
    "    \"Film-Noir\", \"Horror\", \"Musical\", \"Mystery\", \"Romance\", \"Sci-Fi\",\n",
    "    \"Thriller\", \"War\", \"Western\",\n",
    "]\n",
    "\n",
    "movie_titles = item_features[[\"movie_id\", \"movie_title\"]]\n",
    "\n",
    "item_features[\"release_date\"] = pd.to_datetime(item_features[\"release_date\"])\n",
    "item_features = item_features.drop(columns = [\"movie_title\", \"video_release_date\", \"IMDb_URL\", \"unknown\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "single-geology",
   "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>movie_id</th>\n",
       "      <th>release_date</th>\n",
       "      <th>Action</th>\n",
       "      <th>Adventure</th>\n",
       "      <th>Animation</th>\n",
       "      <th>Children's</th>\n",
       "      <th>Comedy</th>\n",
       "      <th>Crime</th>\n",
       "      <th>Documentary</th>\n",
       "      <th>Drama</th>\n",
       "      <th>Fantasy</th>\n",
       "      <th>Film-Noir</th>\n",
       "      <th>Horror</th>\n",
       "      <th>Musical</th>\n",
       "      <th>Mystery</th>\n",
       "      <th>Romance</th>\n",
       "      <th>Sci-Fi</th>\n",
       "      <th>Thriller</th>\n",
       "      <th>War</th>\n",
       "      <th>Western</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>1995-01-01</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>1995-01-01</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>1995-01-01</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>1995-01-01</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>1995-01-01</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1677</th>\n",
       "      <td>1678</td>\n",
       "      <td>1998-02-06</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1678</th>\n",
       "      <td>1679</td>\n",
       "      <td>1998-02-06</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1679</th>\n",
       "      <td>1680</td>\n",
       "      <td>1998-01-01</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1680</th>\n",
       "      <td>1681</td>\n",
       "      <td>1994-01-01</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1681</th>\n",
       "      <td>1682</td>\n",
       "      <td>1996-03-08</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1682 rows × 20 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      movie_id release_date  Action  Adventure  Animation  Children's  Comedy  \\\n",
       "0            1   1995-01-01       0          0          1           1       1   \n",
       "1            2   1995-01-01       1          1          0           0       0   \n",
       "2            3   1995-01-01       0          0          0           0       0   \n",
       "3            4   1995-01-01       1          0          0           0       1   \n",
       "4            5   1995-01-01       0          0          0           0       0   \n",
       "...        ...          ...     ...        ...        ...         ...     ...   \n",
       "1677      1678   1998-02-06       0          0          0           0       0   \n",
       "1678      1679   1998-02-06       0          0          0           0       0   \n",
       "1679      1680   1998-01-01       0          0          0           0       0   \n",
       "1680      1681   1994-01-01       0          0          0           0       1   \n",
       "1681      1682   1996-03-08       0          0          0           0       0   \n",
       "\n",
       "      Crime  Documentary  Drama  Fantasy  Film-Noir  Horror  Musical  Mystery  \\\n",
       "0         0            0      0        0          0       0        0        0   \n",
       "1         0            0      0        0          0       0        0        0   \n",
       "2         0            0      0        0          0       0        0        0   \n",
       "3         0            0      1        0          0       0        0        0   \n",
       "4         1            0      1        0          0       0        0        0   \n",
       "...     ...          ...    ...      ...        ...     ...      ...      ...   \n",
       "1677      0            0      1        0          0       0        0        0   \n",
       "1678      0            0      0        0          0       0        0        0   \n",
       "1679      0            0      1        0          0       0        0        0   \n",
       "1680      0            0      0        0          0       0        0        0   \n",
       "1681      0            0      1        0          0       0        0        0   \n",
       "\n",
       "      Romance  Sci-Fi  Thriller  War  Western  \n",
       "0           0       0         0    0        0  \n",
       "1           0       0         1    0        0  \n",
       "2           0       0         1    0        0  \n",
       "3           0       0         0    0        0  \n",
       "4           0       0         1    0        0  \n",
       "...       ...     ...       ...  ...      ...  \n",
       "1677        0       0         0    0        0  \n",
       "1678        1       0         1    0        0  \n",
       "1679        1       0         0    0        0  \n",
       "1680        0       0         0    0        0  \n",
       "1681        0       0         0    0        0  \n",
       "\n",
       "[1682 rows x 20 columns]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "item_features"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "banned-commercial",
   "metadata": {},
   "source": [
    "# Feature engineering"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "wanted-ribbon",
   "metadata": {},
   "source": [
    "LightFM is able to handle either all binary features or all continuous features.\n",
    "In our data set there are a lot of categorical / binary features, so we will go with all binary features.\n",
    "\n",
    "The following steps are done:\n",
    "* Remove continuous features\n",
    "* Remove zipcodes, that would be too granular feature\n",
    "* Transform categorical into dummy variables"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "violent-apple",
   "metadata": {},
   "outputs": [],
   "source": [
    "user_features = user_features.drop(columns=[\"age\", \"zip_code\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "legal-resident",
   "metadata": {},
   "outputs": [],
   "source": [
    "user_features = user_features \\\n",
    "    .drop(columns=[\"gender\", \"occupation\"]) \\\n",
    "    .join(\n",
    "        pd.get_dummies(user_features[[\"gender\", \"occupation\"]])\n",
    "    )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "found-float",
   "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>user_id</th>\n",
       "      <th>gender_F</th>\n",
       "      <th>gender_M</th>\n",
       "      <th>occupation_administrator</th>\n",
       "      <th>occupation_artist</th>\n",
       "      <th>occupation_doctor</th>\n",
       "      <th>occupation_educator</th>\n",
       "      <th>occupation_engineer</th>\n",
       "      <th>occupation_entertainment</th>\n",
       "      <th>occupation_executive</th>\n",
       "      <th>...</th>\n",
       "      <th>occupation_marketing</th>\n",
       "      <th>occupation_none</th>\n",
       "      <th>occupation_other</th>\n",
       "      <th>occupation_programmer</th>\n",
       "      <th>occupation_retired</th>\n",
       "      <th>occupation_salesman</th>\n",
       "      <th>occupation_scientist</th>\n",
       "      <th>occupation_student</th>\n",
       "      <th>occupation_technician</th>\n",
       "      <th>occupation_writer</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>938</th>\n",
       "      <td>939</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>939</th>\n",
       "      <td>940</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>940</th>\n",
       "      <td>941</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>941</th>\n",
       "      <td>942</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>942</th>\n",
       "      <td>943</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>943 rows × 24 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     user_id  gender_F  gender_M  occupation_administrator  occupation_artist  \\\n",
       "0          1         0         1                         0                  0   \n",
       "1          2         1         0                         0                  0   \n",
       "2          3         0         1                         0                  0   \n",
       "3          4         0         1                         0                  0   \n",
       "4          5         1         0                         0                  0   \n",
       "..       ...       ...       ...                       ...                ...   \n",
       "938      939         1         0                         0                  0   \n",
       "939      940         0         1                         1                  0   \n",
       "940      941         0         1                         0                  0   \n",
       "941      942         1         0                         0                  0   \n",
       "942      943         0         1                         0                  0   \n",
       "\n",
       "     occupation_doctor  occupation_educator  occupation_engineer  \\\n",
       "0                    0                    0                    0   \n",
       "1                    0                    0                    0   \n",
       "2                    0                    0                    0   \n",
       "3                    0                    0                    0   \n",
       "4                    0                    0                    0   \n",
       "..                 ...                  ...                  ...   \n",
       "938                  0                    0                    0   \n",
       "939                  0                    0                    0   \n",
       "940                  0                    0                    0   \n",
       "941                  0                    0                    0   \n",
       "942                  0                    0                    0   \n",
       "\n",
       "     occupation_entertainment  occupation_executive  ...  \\\n",
       "0                           0                     0  ...   \n",
       "1                           0                     0  ...   \n",
       "2                           0                     0  ...   \n",
       "3                           0                     0  ...   \n",
       "4                           0                     0  ...   \n",
       "..                        ...                   ...  ...   \n",
       "938                         0                     0  ...   \n",
       "939                         0                     0  ...   \n",
       "940                         0                     0  ...   \n",
       "941                         0                     0  ...   \n",
       "942                         0                     0  ...   \n",
       "\n",
       "     occupation_marketing  occupation_none  occupation_other  \\\n",
       "0                       0                0                 0   \n",
       "1                       0                0                 1   \n",
       "2                       0                0                 0   \n",
       "3                       0                0                 0   \n",
       "4                       0                0                 1   \n",
       "..                    ...              ...               ...   \n",
       "938                     0                0                 0   \n",
       "939                     0                0                 0   \n",
       "940                     0                0                 0   \n",
       "941                     0                0                 0   \n",
       "942                     0                0                 0   \n",
       "\n",
       "     occupation_programmer  occupation_retired  occupation_salesman  \\\n",
       "0                        0                   0                    0   \n",
       "1                        0                   0                    0   \n",
       "2                        0                   0                    0   \n",
       "3                        0                   0                    0   \n",
       "4                        0                   0                    0   \n",
       "..                     ...                 ...                  ...   \n",
       "938                      0                   0                    0   \n",
       "939                      0                   0                    0   \n",
       "940                      0                   0                    0   \n",
       "941                      0                   0                    0   \n",
       "942                      0                   0                    0   \n",
       "\n",
       "     occupation_scientist  occupation_student  occupation_technician  \\\n",
       "0                       0                   0                      1   \n",
       "1                       0                   0                      0   \n",
       "2                       0                   0                      0   \n",
       "3                       0                   0                      1   \n",
       "4                       0                   0                      0   \n",
       "..                    ...                 ...                    ...   \n",
       "938                     0                   1                      0   \n",
       "939                     0                   0                      0   \n",
       "940                     0                   1                      0   \n",
       "941                     0                   0                      0   \n",
       "942                     0                   1                      0   \n",
       "\n",
       "     occupation_writer  \n",
       "0                    0  \n",
       "1                    0  \n",
       "2                    1  \n",
       "3                    0  \n",
       "4                    0  \n",
       "..                 ...  \n",
       "938                  0  \n",
       "939                  0  \n",
       "940                  0  \n",
       "941                  0  \n",
       "942                  0  \n",
       "\n",
       "[943 rows x 24 columns]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "user_features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "informational-alabama",
   "metadata": {},
   "outputs": [],
   "source": [
    "item_features = item_features.drop(columns=[\"release_date\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "advanced-sweden",
   "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>movie_id</th>\n",
       "      <th>Action</th>\n",
       "      <th>Adventure</th>\n",
       "      <th>Animation</th>\n",
       "      <th>Children's</th>\n",
       "      <th>Comedy</th>\n",
       "      <th>Crime</th>\n",
       "      <th>Documentary</th>\n",
       "      <th>Drama</th>\n",
       "      <th>Fantasy</th>\n",
       "      <th>Film-Noir</th>\n",
       "      <th>Horror</th>\n",
       "      <th>Musical</th>\n",
       "      <th>Mystery</th>\n",
       "      <th>Romance</th>\n",
       "      <th>Sci-Fi</th>\n",
       "      <th>Thriller</th>\n",
       "      <th>War</th>\n",
       "      <th>Western</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1677</th>\n",
       "      <td>1678</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1678</th>\n",
       "      <td>1679</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1679</th>\n",
       "      <td>1680</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1680</th>\n",
       "      <td>1681</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1681</th>\n",
       "      <td>1682</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1682 rows × 19 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      movie_id  Action  Adventure  Animation  Children's  Comedy  Crime  \\\n",
       "0            1       0          0          1           1       1      0   \n",
       "1            2       1          1          0           0       0      0   \n",
       "2            3       0          0          0           0       0      0   \n",
       "3            4       1          0          0           0       1      0   \n",
       "4            5       0          0          0           0       0      1   \n",
       "...        ...     ...        ...        ...         ...     ...    ...   \n",
       "1677      1678       0          0          0           0       0      0   \n",
       "1678      1679       0          0          0           0       0      0   \n",
       "1679      1680       0          0          0           0       0      0   \n",
       "1680      1681       0          0          0           0       1      0   \n",
       "1681      1682       0          0          0           0       0      0   \n",
       "\n",
       "      Documentary  Drama  Fantasy  Film-Noir  Horror  Musical  Mystery  \\\n",
       "0               0      0        0          0       0        0        0   \n",
       "1               0      0        0          0       0        0        0   \n",
       "2               0      0        0          0       0        0        0   \n",
       "3               0      1        0          0       0        0        0   \n",
       "4               0      1        0          0       0        0        0   \n",
       "...           ...    ...      ...        ...     ...      ...      ...   \n",
       "1677            0      1        0          0       0        0        0   \n",
       "1678            0      0        0          0       0        0        0   \n",
       "1679            0      1        0          0       0        0        0   \n",
       "1680            0      0        0          0       0        0        0   \n",
       "1681            0      1        0          0       0        0        0   \n",
       "\n",
       "      Romance  Sci-Fi  Thriller  War  Western  \n",
       "0           0       0         0    0        0  \n",
       "1           0       0         1    0        0  \n",
       "2           0       0         1    0        0  \n",
       "3           0       0         0    0        0  \n",
       "4           0       0         1    0        0  \n",
       "...       ...     ...       ...  ...      ...  \n",
       "1677        0       0         0    0        0  \n",
       "1678        1       0         1    0        0  \n",
       "1679        1       0         0    0        0  \n",
       "1680        0       0         0    0        0  \n",
       "1681        0       0         0    0        0  \n",
       "\n",
       "[1682 rows x 19 columns]"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "item_features"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "alone-accountability",
   "metadata": {},
   "source": [
    "# Split to current / new, train / validation sets"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "raised-guinea",
   "metadata": {},
   "source": [
    "In this notebook we simulate the following scenario. We train the complete model at a certain point of time, we push this model to production.\n",
    "But later on the users make new interactions and we also want to incorporate those to the model.\n",
    "\n",
    "\"Current\" set will represent the interactions known at the time of training the complete model.\n",
    "\"New\" set will represent the new interactions, that we also want to incorporate to the model."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "threaded-landscape",
   "metadata": {},
   "source": [
    "## Interactions"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "outstanding-decrease",
   "metadata": {},
   "source": [
    "### Split current, new on time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "varying-karen",
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_split_time(interactions, current_ratio):\n",
    "    split_time = interactions \\\n",
    "        .sort_values(\"time\") \\\n",
    "        .reset_index() \\\n",
    "        .iloc[int(interactions.shape[0] * current_ratio)] \\\n",
    "        [\"time\"]\n",
    "    return split_time\n",
    "\n",
    "def current_new_split(interactions, current_ratio):\n",
    "    split_time = get_split_time(interactions, current_ratio)\n",
    "\n",
    "    interactions_current = interactions[interactions[\"time\"] < split_time]\n",
    "    interactions_new = interactions[interactions[\"time\"] >= split_time]\n",
    "    \n",
    "    return interactions_current, interactions_new\n",
    "    \n",
    "interactions_current, interactions_new = current_new_split(interactions, 0.9)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "basic-tradition",
   "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>user_id</th>\n",
       "      <th>movie_id</th>\n",
       "      <th>rating</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>196</td>\n",
       "      <td>242</td>\n",
       "      <td>3</td>\n",
       "      <td>1997-12-04 15:55:49</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>298</td>\n",
       "      <td>474</td>\n",
       "      <td>4</td>\n",
       "      <td>1998-01-07 14:20:06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>305</td>\n",
       "      <td>451</td>\n",
       "      <td>3</td>\n",
       "      <td>1998-02-01 09:20:17</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>6</td>\n",
       "      <td>86</td>\n",
       "      <td>3</td>\n",
       "      <td>1997-12-31 21:16:53</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>286</td>\n",
       "      <td>1014</td>\n",
       "      <td>5</td>\n",
       "      <td>1997-11-17 15:38:45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99992</th>\n",
       "      <td>721</td>\n",
       "      <td>262</td>\n",
       "      <td>3</td>\n",
       "      <td>1997-10-18 01:14:45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99994</th>\n",
       "      <td>378</td>\n",
       "      <td>78</td>\n",
       "      <td>3</td>\n",
       "      <td>1997-11-20 20:16:16</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99995</th>\n",
       "      <td>880</td>\n",
       "      <td>476</td>\n",
       "      <td>3</td>\n",
       "      <td>1997-11-22 05:10:44</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99996</th>\n",
       "      <td>716</td>\n",
       "      <td>204</td>\n",
       "      <td>5</td>\n",
       "      <td>1997-11-17 19:39:03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99999</th>\n",
       "      <td>12</td>\n",
       "      <td>203</td>\n",
       "      <td>3</td>\n",
       "      <td>1997-11-19 17:13:03</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>74266 rows × 4 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       user_id  movie_id  rating                time\n",
       "0          196       242       3 1997-12-04 15:55:49\n",
       "5          298       474       4 1998-01-07 14:20:06\n",
       "8          305       451       3 1998-02-01 09:20:17\n",
       "9            6        86       3 1997-12-31 21:16:53\n",
       "11         286      1014       5 1997-11-17 15:38:45\n",
       "...        ...       ...     ...                 ...\n",
       "99992      721       262       3 1997-10-18 01:14:45\n",
       "99994      378        78       3 1997-11-20 20:16:16\n",
       "99995      880       476       3 1997-11-22 05:10:44\n",
       "99996      716       204       5 1997-11-17 19:39:03\n",
       "99999       12       203       3 1997-11-19 17:13:03\n",
       "\n",
       "[74266 rows x 4 columns]"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "interactions_current"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "single-folder",
   "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>user_id</th>\n",
       "      <th>movie_id</th>\n",
       "      <th>rating</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>186</td>\n",
       "      <td>302</td>\n",
       "      <td>3</td>\n",
       "      <td>1998-04-04 19:22:22</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>253</td>\n",
       "      <td>465</td>\n",
       "      <td>5</td>\n",
       "      <td>1998-04-03 18:34:27</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>167</td>\n",
       "      <td>486</td>\n",
       "      <td>4</td>\n",
       "      <td>1998-04-16 14:54:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>38</td>\n",
       "      <td>95</td>\n",
       "      <td>5</td>\n",
       "      <td>1998-04-13 01:14:54</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>69</th>\n",
       "      <td>223</td>\n",
       "      <td>274</td>\n",
       "      <td>4</td>\n",
       "      <td>1998-04-02 20:48:14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99921</th>\n",
       "      <td>659</td>\n",
       "      <td>657</td>\n",
       "      <td>5</td>\n",
       "      <td>1998-03-31 22:39:25</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99942</th>\n",
       "      <td>363</td>\n",
       "      <td>181</td>\n",
       "      <td>5</td>\n",
       "      <td>1998-04-02 05:26:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99953</th>\n",
       "      <td>655</td>\n",
       "      <td>913</td>\n",
       "      <td>4</td>\n",
       "      <td>1998-04-05 23:05:21</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99988</th>\n",
       "      <td>421</td>\n",
       "      <td>498</td>\n",
       "      <td>4</td>\n",
       "      <td>1998-04-10 20:49:04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99991</th>\n",
       "      <td>676</td>\n",
       "      <td>538</td>\n",
       "      <td>4</td>\n",
       "      <td>1998-04-16 00:10:37</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>8254 rows × 4 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       user_id  movie_id  rating                time\n",
       "1          186       302       3 1998-04-04 19:22:22\n",
       "7          253       465       5 1998-04-03 18:34:27\n",
       "21         167       486       4 1998-04-16 14:54:12\n",
       "26          38        95       5 1998-04-13 01:14:54\n",
       "69         223       274       4 1998-04-02 20:48:14\n",
       "...        ...       ...     ...                 ...\n",
       "99921      659       657       5 1998-03-31 22:39:25\n",
       "99942      363       181       5 1998-04-02 05:26:23\n",
       "99953      655       913       4 1998-04-05 23:05:21\n",
       "99988      421       498       4 1998-04-10 20:49:04\n",
       "99991      676       538       4 1998-04-16 00:10:37\n",
       "\n",
       "[8254 rows x 4 columns]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "interactions_new"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "sapphire-vampire",
   "metadata": {},
   "source": [
    "Let's have a look at how many new users and items are in the \"new\" set. New user / item means that they weren't present in the \"current\" set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "extended-chorus",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of new users in new set: 76\n",
      "Number of current users in new set: 86\n"
     ]
    }
   ],
   "source": [
    "new_users = np.setdiff1d(interactions_new[\"user_id\"], interactions_current[\"user_id\"])\n",
    "old_users = np.intersect1d(interactions_new[\"user_id\"], interactions_current[\"user_id\"])\n",
    "\n",
    "print(\"Number of new users in new set: {}\".format(len(new_users)))\n",
    "print(\"Number of current users in new set: {}\".format(len(old_users)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "virtual-treat",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of new items in new set: 49\n",
      "Number of current items in new set: 1167\n"
     ]
    }
   ],
   "source": [
    "new_items = np.setdiff1d(interactions_new[\"movie_id\"], interactions_current[\"movie_id\"])\n",
    "old_items = np.intersect1d(interactions_new[\"movie_id\"], interactions_current[\"movie_id\"])\n",
    "\n",
    "print(\"Number of new items in new set: {}\".format(len(new_items)))\n",
    "print(\"Number of current items in new set: {}\".format(len(old_items)))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "together-ticket",
   "metadata": {},
   "source": [
    "Let's visualize the interactions in time, along with the split."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "swedish-bloom",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "users_first_interactions = interactions \\\n",
    "    .groupby(\"user_id\") \\\n",
    "    .agg({\"time\": \"min\"}) \\\n",
    "    .reset_index() \\\n",
    "    .rename(columns={\"time\": \"first_time\"})\n",
    "users_first_interactions = users_first_interactions.sort_values(\"first_time\")\n",
    "users_first_interactions[\"first_rank\"] = range(users_first_interactions.shape[0])\n",
    "\n",
    "interactions_first_time = interactions.merge(users_first_interactions, on=\"user_id\")\n",
    "\n",
    "split_time = get_split_time(interactions, 0.9)\n",
    "plt.figure(figsize=(8, 8))\n",
    "plt.scatter(\n",
    "    interactions_first_time[\"time\"],\n",
    "    interactions_first_time[\"first_rank\"],\n",
    "    alpha=0.1\n",
    ")\n",
    "plt.plot(\n",
    "    [split_time, split_time],\n",
    "    [interactions_first_time[\"first_rank\"].min(), interactions_first_time[\"first_rank\"].max()],\n",
    "    color=\"red\"\n",
    ")\n",
    "plt.title(\n",
    "    \"User interactions in time\"\n",
    ")\n",
    "plt.xlabel(\"Time\")\n",
    "plt.ylabel(\"Users\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ruled-retreat",
   "metadata": {},
   "source": [
    "Users are ordered by their first interaction. Every dot represents an interaction, and the red line represents the 'current', 'new' split.\n",
    "\n",
    "This explains why we see new users in the \"new\" data set."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "instant-monster",
   "metadata": {},
   "source": [
    "### Split train, validation randomly"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "numeric-accessory",
   "metadata": {},
   "outputs": [],
   "source": [
    "def train_test_split_randomly(data, test_size):\n",
    "    \"\"\"\n",
    "    Randomly splits interactions to train and test.\n",
    "    \n",
    "    Makes sure there is at least 1 interaction for every user and item in the train set.\n",
    "    \"\"\"\n",
    "    \n",
    "    # Split randomly\n",
    "    random_numbers = np.random.uniform(size=data.shape[0])\n",
    "    test_mask = random_numbers <= test_size\n",
    "    data_train = data[~test_mask]\n",
    "    data_test = data[test_mask]\n",
    "    \n",
    "    # Make sure there is at least 1 interaction for every user and item in the train set\n",
    "    for id_name in [\"user_id\", \"movie_id\"]:\n",
    "        training_ids = data_train[id_name].unique()\n",
    "        for id_ in data[id_name].unique():\n",
    "            if id_ not in training_ids:\n",
    "                row_to_training = data_test[data_test[id_name] == id_].sample(n=1)\n",
    "                data_test = data_test.drop(index=row_to_training.index)\n",
    "                data_train = data_train.append(row_to_training)\n",
    "                \n",
    "    return data_train, data_test\n",
    "\n",
    "\n",
    "interactions_current_train, interactions_current_val = train_test_split_randomly(interactions_current, 0.3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "false-auditor",
   "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>user_id</th>\n",
       "      <th>movie_id</th>\n",
       "      <th>rating</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>196</td>\n",
       "      <td>242</td>\n",
       "      <td>3</td>\n",
       "      <td>1997-12-04 15:55:49</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>286</td>\n",
       "      <td>1014</td>\n",
       "      <td>5</td>\n",
       "      <td>1997-11-17 15:38:45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>210</td>\n",
       "      <td>40</td>\n",
       "      <td>3</td>\n",
       "      <td>1998-03-27 21:59:54</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>224</td>\n",
       "      <td>29</td>\n",
       "      <td>3</td>\n",
       "      <td>1998-02-21 23:40:57</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>122</td>\n",
       "      <td>387</td>\n",
       "      <td>5</td>\n",
       "      <td>1997-11-11 17:47:39</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>96166</th>\n",
       "      <td>870</td>\n",
       "      <td>1664</td>\n",
       "      <td>4</td>\n",
       "      <td>1998-03-16 14:08:42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>97122</th>\n",
       "      <td>877</td>\n",
       "      <td>1402</td>\n",
       "      <td>4</td>\n",
       "      <td>1997-12-21 04:09:46</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>98427</th>\n",
       "      <td>655</td>\n",
       "      <td>1637</td>\n",
       "      <td>3</td>\n",
       "      <td>1998-03-04 04:04:15</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>98640</th>\n",
       "      <td>655</td>\n",
       "      <td>1630</td>\n",
       "      <td>3</td>\n",
       "      <td>1998-02-14 03:58:55</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>98946</th>\n",
       "      <td>804</td>\n",
       "      <td>1177</td>\n",
       "      <td>3</td>\n",
       "      <td>1997-11-13 18:39:50</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>52154 rows × 4 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       user_id  movie_id  rating                time\n",
       "0          196       242       3 1997-12-04 15:55:49\n",
       "11         286      1014       5 1997-11-17 15:38:45\n",
       "13         210        40       3 1998-03-27 21:59:54\n",
       "14         224        29       3 1998-02-21 23:40:57\n",
       "16         122       387       5 1997-11-11 17:47:39\n",
       "...        ...       ...     ...                 ...\n",
       "96166      870      1664       4 1998-03-16 14:08:42\n",
       "97122      877      1402       4 1997-12-21 04:09:46\n",
       "98427      655      1637       3 1998-03-04 04:04:15\n",
       "98640      655      1630       3 1998-02-14 03:58:55\n",
       "98946      804      1177       3 1997-11-13 18:39:50\n",
       "\n",
       "[52154 rows x 4 columns]"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "interactions_current_train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "voluntary-means",
   "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>user_id</th>\n",
       "      <th>movie_id</th>\n",
       "      <th>rating</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>298</td>\n",
       "      <td>474</td>\n",
       "      <td>4</td>\n",
       "      <td>1998-01-07 14:20:06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>305</td>\n",
       "      <td>451</td>\n",
       "      <td>3</td>\n",
       "      <td>1998-02-01 09:20:17</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>6</td>\n",
       "      <td>86</td>\n",
       "      <td>3</td>\n",
       "      <td>1997-12-31 21:16:53</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>200</td>\n",
       "      <td>222</td>\n",
       "      <td>5</td>\n",
       "      <td>1997-10-05 09:05:40</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>303</td>\n",
       "      <td>785</td>\n",
       "      <td>3</td>\n",
       "      <td>1997-11-14 05:28:38</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99963</th>\n",
       "      <td>429</td>\n",
       "      <td>199</td>\n",
       "      <td>5</td>\n",
       "      <td>1997-12-17 19:13:26</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99967</th>\n",
       "      <td>897</td>\n",
       "      <td>369</td>\n",
       "      <td>4</td>\n",
       "      <td>1997-11-20 02:41:53</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99972</th>\n",
       "      <td>721</td>\n",
       "      <td>874</td>\n",
       "      <td>3</td>\n",
       "      <td>1997-10-18 01:17:27</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99974</th>\n",
       "      <td>764</td>\n",
       "      <td>596</td>\n",
       "      <td>3</td>\n",
       "      <td>1997-10-07 16:50:46</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99978</th>\n",
       "      <td>113</td>\n",
       "      <td>975</td>\n",
       "      <td>5</td>\n",
       "      <td>1997-10-04 03:40:24</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>22112 rows × 4 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       user_id  movie_id  rating                time\n",
       "5          298       474       4 1998-01-07 14:20:06\n",
       "8          305       451       3 1998-02-01 09:20:17\n",
       "9            6        86       3 1997-12-31 21:16:53\n",
       "12         200       222       5 1997-10-05 09:05:40\n",
       "15         303       785       3 1997-11-14 05:28:38\n",
       "...        ...       ...     ...                 ...\n",
       "99963      429       199       5 1997-12-17 19:13:26\n",
       "99967      897       369       4 1997-11-20 02:41:53\n",
       "99972      721       874       3 1997-10-18 01:17:27\n",
       "99974      764       596       3 1997-10-07 16:50:46\n",
       "99978      113       975       5 1997-10-04 03:40:24\n",
       "\n",
       "[22112 rows x 4 columns]"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "interactions_current_val"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "searching-upgrade",
   "metadata": {},
   "source": [
    "## User, item features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "floating-edition",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of current users: 867\n",
      "Number of new users: 162\n"
     ]
    }
   ],
   "source": [
    "user_features_current = user_features[user_features[\"user_id\"].isin(interactions_current[\"user_id\"])]\n",
    "user_features_new = user_features[user_features[\"user_id\"].isin(interactions_new[\"user_id\"])]\n",
    "\n",
    "print(\"Number of current users: {}\".format(len(user_features_current)))\n",
    "print(\"Number of new users: {}\".format(len(user_features_new)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "committed-controversy",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of current items: 1525\n",
      "Number of new items: 1216\n"
     ]
    }
   ],
   "source": [
    "item_features_current = item_features[item_features[\"movie_id\"].isin(interactions_current[\"movie_id\"])]\n",
    "item_features_new = item_features[item_features[\"movie_id\"].isin(interactions_new[\"movie_id\"])]\n",
    "\n",
    "print(\"Number of current items: {}\".format(len(item_features_current)))\n",
    "print(\"Number of new items: {}\".format(len(item_features_new)))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "damaged-adapter",
   "metadata": {},
   "source": [
    "# Transform to LightFM format"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "applied-software",
   "metadata": {},
   "source": [
    "## Dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "flexible-league",
   "metadata": {},
   "source": [
    "We intentionally initialize the dataset with more users / items that there are in the \"current\" data set.\n",
    "The reason is that the model size can't be changed without retraining the complete model, but we want to be\n",
    "able to incorporate interactions from potential new users / items. This way there is room for them."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "environmental-equivalent",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Max user id: 943\n",
      "Max movie id: 1682\n"
     ]
    }
   ],
   "source": [
    "print(\"Max user id: {}\".format(interactions_current['user_id'].max()))\n",
    "print(\"Max movie id: {}\".format(interactions_current['movie_id'].max()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "crazy-bosnia",
   "metadata": {},
   "outputs": [],
   "source": [
    "user_ids_buffered = range(1000)\n",
    "item_ids_buffered = range(1700)\n",
    "\n",
    "user_feature_names = user_features.columns.tolist()\n",
    "user_feature_names.remove(\"user_id\")\n",
    "\n",
    "item_feature_names = item_features.columns.tolist()\n",
    "item_feature_names.remove(\"movie_id\")\n",
    "\n",
    "dataset = Dataset()\n",
    "dataset.fit(\n",
    "    users=user_ids_buffered,\n",
    "    items=item_ids_buffered,\n",
    "    user_features=user_feature_names,\n",
    "    item_features=item_feature_names\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "filled-characterization",
   "metadata": {},
   "source": [
    "## Interaction matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "continental-louis",
   "metadata": {},
   "outputs": [],
   "source": [
    "def transform_interactions(interactions):\n",
    "    \"\"\"\n",
    "    Transforms interactions data frame to the format build_interactions needs it.\n",
    "    \n",
    "    Returns iterable of tuples (user_id, movie_id, weight).\n",
    "    \"\"\"\n",
    "    return interactions[[\"user_id\", \"movie_id\", \"rating\"]].itertuples(index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "visible-attraction",
   "metadata": {},
   "outputs": [],
   "source": [
    "interaction_matrix_current, _ = dataset.build_interactions(\n",
    "    transform_interactions(interactions_current))\n",
    "\n",
    "interaction_matrix_current_train, _ = dataset.build_interactions(\n",
    "    transform_interactions(interactions_current_train))\n",
    "\n",
    "interaction_matrix_current_val, _ = dataset.build_interactions(\n",
    "    transform_interactions(interactions_current_val))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "social-headline",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1000, 1700)\n",
      "(1000, 1700)\n",
      "(1000, 1700)\n"
     ]
    }
   ],
   "source": [
    "print(interaction_matrix_current.shape)\n",
    "print(interaction_matrix_current_train.shape)\n",
    "print(interaction_matrix_current_val.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "loving-program",
   "metadata": {},
   "source": [
    "## User, item features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "written-enhancement",
   "metadata": {},
   "outputs": [],
   "source": [
    "def transform_features(features, id_name):\n",
    "    \"\"\"\n",
    "    Transforms features data frame to the format build_user_features / build_item_features needs it.\n",
    "    \n",
    "    Returns iterable of (id, [list of feature names that has value 1]).\n",
    "    \"\"\"\n",
    "    transformed_features = []\n",
    "    for row in features.to_dict(orient=\"records\"):\n",
    "        id_value = row[id_name]\n",
    "        del row[id_name]\n",
    "        feature_names = [key for key, value in row.items() if value == 1]\n",
    "        transformed_features.append(\n",
    "            (id_value, feature_names)\n",
    "        )\n",
    "    return transformed_features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "bizarre-separate",
   "metadata": {},
   "outputs": [],
   "source": [
    "user_features_matrix_current = dataset.build_user_features(\n",
    "    transform_features(user_features_current, \"user_id\"))\n",
    "\n",
    "item_features_matrix_current = dataset.build_item_features(\n",
    "    transform_features(item_features_current, \"movie_id\"))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "color-badge",
   "metadata": {},
   "source": [
    "# Train"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fantastic-birmingham",
   "metadata": {},
   "source": [
    "## Model definition"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "exempt-bronze",
   "metadata": {},
   "outputs": [],
   "source": [
    "model = LightFM(loss=\"warp\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "false-tuner",
   "metadata": {},
   "source": [
    "## Training and hyperparameter tuning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "collected-console",
   "metadata": {},
   "outputs": [],
   "source": [
    "class Evaluator():\n",
    "    \"\"\"\n",
    "    Helper class for model evaluation\n",
    "    \n",
    "    Stores the number of epochs and corresponding performance metrics in a data frame.\n",
    "    Returns the number of epochs with the best validation ROC AUC.\n",
    "    Prepared to handle multiple runs of the same hyperparameters, and averages up the multiple runs.\n",
    "    \"\"\"\n",
    "    \n",
    "    def __init__(self):\n",
    "        self.df_evaluation = pd.DataFrame()\n",
    "    def evaluate(self, epochs, model, interaction_matrix_train, interaction_matrix_val, user_features_matrix, item_features_matrix):\n",
    "            features_current = {\n",
    "                \"user_features\": user_features_matrix,\n",
    "                \"item_features\": item_features_matrix\n",
    "            }\n",
    "            evaluation_row = pd.DataFrame({\n",
    "                \"epochs\": epochs,\n",
    "                \"train_precision\": precision_at_k(model, interaction_matrix_train, k=10, **features_current).mean(),\n",
    "                \"validation_precision\": precision_at_k(model, interaction_matrix_val, k=10, **features_current).mean(),\n",
    "                \"train_auc\": auc_score(model, interaction_matrix_train, **features_current).mean(),\n",
    "                \"validation_auc\": auc_score(model, interaction_matrix_val, **features_current).mean()\n",
    "                },\n",
    "                index=[0]\n",
    "            )\n",
    "            self.df_evaluation = self.df_evaluation.append(evaluation_row, ignore_index=True)\n",
    "    def get_best_epochs(self):\n",
    "        df_evaluation_agg = self.df_evaluation.groupby(\"epochs\").agg(\"mean\").reset_index()\n",
    "        best_epochs =  df_evaluation_agg[\n",
    "            df_evaluation_agg[\"validation_auc\"] == df_evaluation_agg[\"validation_auc\"].max()\n",
    "        ][\"epochs\"].iloc[0]\n",
    "        return best_epochs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "sorted-handle",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epochs: 100\n",
      "Epochs: 200\n",
      "Epochs: 300\n",
      "Epochs: 500\n",
      "Epochs: 1000\n",
      "Epochs: 2000\n"
     ]
    }
   ],
   "source": [
    "evaluator = Evaluator()\n",
    "for epochs in [100, 200, 300, 500, 1000, 2000]:\n",
    "    print(f\"Epochs: {epochs}\")\n",
    "    \n",
    "    model.fit(\n",
    "        interaction_matrix_current_train,\n",
    "        epochs=epochs,\n",
    "        user_features=user_features_matrix_current,\n",
    "        item_features=item_features_matrix_current\n",
    "    )\n",
    "    \n",
    "    evaluator.evaluate(\n",
    "        epochs, model,\n",
    "        interaction_matrix_current_train, interaction_matrix_current_val,\n",
    "        user_features_matrix_current, item_features_matrix_current)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "functioning-parts",
   "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>epochs</th>\n",
       "      <th>train_precision</th>\n",
       "      <th>validation_precision</th>\n",
       "      <th>train_auc</th>\n",
       "      <th>validation_auc</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>100</td>\n",
       "      <td>0.413495</td>\n",
       "      <td>0.146420</td>\n",
       "      <td>0.907756</td>\n",
       "      <td>0.883156</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>200</td>\n",
       "      <td>0.457670</td>\n",
       "      <td>0.160277</td>\n",
       "      <td>0.926467</td>\n",
       "      <td>0.897800</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>300</td>\n",
       "      <td>0.497347</td>\n",
       "      <td>0.167090</td>\n",
       "      <td>0.938702</td>\n",
       "      <td>0.907911</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>500</td>\n",
       "      <td>0.514072</td>\n",
       "      <td>0.168129</td>\n",
       "      <td>0.946965</td>\n",
       "      <td>0.914250</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1000</td>\n",
       "      <td>0.537716</td>\n",
       "      <td>0.167090</td>\n",
       "      <td>0.953773</td>\n",
       "      <td>0.918350</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>2000</td>\n",
       "      <td>0.544752</td>\n",
       "      <td>0.165473</td>\n",
       "      <td>0.958380</td>\n",
       "      <td>0.919866</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   epochs  train_precision  validation_precision  train_auc  validation_auc\n",
       "0     100         0.413495              0.146420   0.907756        0.883156\n",
       "1     200         0.457670              0.160277   0.926467        0.897800\n",
       "2     300         0.497347              0.167090   0.938702        0.907911\n",
       "3     500         0.514072              0.168129   0.946965        0.914250\n",
       "4    1000         0.537716              0.167090   0.953773        0.918350\n",
       "5    2000         0.544752              0.165473   0.958380        0.919866"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "evaluator.df_evaluation"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "accomplished-collapse",
   "metadata": {},
   "source": [
    "## Retrain on complete training dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "corrected-handle",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best number of epochs: 2000\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<lightfm.lightfm.LightFM at 0x12e094640>"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "best_epochs = evaluator.get_best_epochs()\n",
    "print(f\"Best number of epochs: {best_epochs}\")\n",
    "\n",
    "model.fit(\n",
    "    interaction_matrix_current,\n",
    "    epochs=best_epochs,\n",
    "    user_features=user_features_matrix_current,\n",
    "    item_features=item_features_matrix_current\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "united-cartridge",
   "metadata": {},
   "source": [
    "# Train and predict for new"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "overhead-campus",
   "metadata": {},
   "source": [
    "## Create matrices"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "residential-nightlife",
   "metadata": {},
   "outputs": [],
   "source": [
    "interaction_matrix_new, _ = dataset.build_interactions(\n",
    "    transform_interactions(interactions_new))\n",
    "\n",
    "user_features_matrix_new = dataset.build_user_features(\n",
    "    transform_features(user_features_new, \"user_id\"))\n",
    "\n",
    "item_features_matrix_new = dataset.build_item_features(\n",
    "    transform_features(item_features_new, \"movie_id\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "retired-northwest",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1000, 1700)\n",
      "(1000, 1023)\n",
      "(1700, 1718)\n"
     ]
    }
   ],
   "source": [
    "print(interaction_matrix_new.shape)\n",
    "print(user_features_matrix_new.shape)\n",
    "print(item_features_matrix_new.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "bibliographic-penalty",
   "metadata": {},
   "outputs": [],
   "source": [
    "user_features_matrix = dataset.build_user_features(\n",
    "    transform_features(user_features, \"user_id\"))\n",
    "\n",
    "item_features_matrix = dataset.build_item_features(\n",
    "    transform_features(item_features, \"movie_id\"))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "irish-workplace",
   "metadata": {},
   "source": [
    "## Fit partial"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "inappropriate-major",
   "metadata": {},
   "source": [
    "Incorporate new interactions to the model. Since the number of new interactions can be small, for every hyperparameter set,\n",
    "the train test split is done multiple times, and the performance is averaged up."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "happy-bacteria",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epochs: 100\n",
      "fold: 0\n",
      "fold: 1\n",
      "fold: 2\n",
      "fold: 3\n",
      "fold: 4\n",
      "Epochs: 500\n",
      "fold: 0\n",
      "fold: 1\n",
      "fold: 2\n",
      "fold: 3\n",
      "fold: 4\n",
      "Epochs: 1000\n",
      "fold: 0\n",
      "fold: 1\n",
      "fold: 2\n",
      "fold: 3\n",
      "fold: 4\n",
      "Epochs: 2000\n",
      "fold: 0\n",
      "fold: 1\n",
      "fold: 2\n",
      "fold: 3\n",
      "fold: 4\n"
     ]
    }
   ],
   "source": [
    "evaluator = Evaluator()\n",
    "for epochs in [100, 500, 1000, 2000]:\n",
    "    print(f\"Epochs: {epochs}\")\n",
    "\n",
    "    num_fold = 5\n",
    "    for fold in range(num_fold):\n",
    "        print(f\"fold: {fold}\")\n",
    "              \n",
    "        interactions_new_train, interactions_new_val = train_test_split_randomly(interactions_new, 1/num_fold)\n",
    "        interaction_matrix_new_train, _ = dataset.build_interactions(\n",
    "            transform_interactions(interactions_new_train))\n",
    "        interaction_matrix_new_val, _ = dataset.build_interactions(\n",
    "            transform_interactions(interactions_new_val))\n",
    "        \n",
    "        model_fold = deepcopy(model)\n",
    "        model_fold.fit_partial(\n",
    "            interaction_matrix_new_train,\n",
    "            user_features=user_features_matrix_new,\n",
    "            item_features=item_features_matrix_new,\n",
    "            epochs=epochs\n",
    "        )\n",
    "\n",
    "        evaluator.evaluate(\n",
    "            epochs, model_fold,\n",
    "            interaction_matrix_new_train, interaction_matrix_new_val,\n",
    "            user_features_matrix_new, item_features_matrix_new)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "possible-disco",
   "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>epochs</th>\n",
       "      <th>train_precision</th>\n",
       "      <th>validation_precision</th>\n",
       "      <th>train_auc</th>\n",
       "      <th>validation_auc</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>100</td>\n",
       "      <td>0.299136</td>\n",
       "      <td>0.081650</td>\n",
       "      <td>0.892203</td>\n",
       "      <td>0.887516</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>500</td>\n",
       "      <td>0.345062</td>\n",
       "      <td>0.077575</td>\n",
       "      <td>0.921617</td>\n",
       "      <td>0.902536</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1000</td>\n",
       "      <td>0.356543</td>\n",
       "      <td>0.076211</td>\n",
       "      <td>0.936391</td>\n",
       "      <td>0.904481</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2000</td>\n",
       "      <td>0.381111</td>\n",
       "      <td>0.073123</td>\n",
       "      <td>0.952694</td>\n",
       "      <td>0.901498</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   epochs  train_precision  validation_precision  train_auc  validation_auc\n",
       "0     100         0.299136              0.081650   0.892203        0.887516\n",
       "1     500         0.345062              0.077575   0.921617        0.902536\n",
       "2    1000         0.356543              0.076211   0.936391        0.904481\n",
       "3    2000         0.381111              0.073123   0.952694        0.901498"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "evaluator.df_evaluation.groupby(\"epochs\").agg(\"mean\").reset_index()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "known-overall",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best number of epochs: 1000\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<lightfm.lightfm.LightFM at 0x12e094640>"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "best_epochs = evaluator.get_best_epochs()\n",
    "print(f\"Best number of epochs: {best_epochs}\")\n",
    "\n",
    "model.fit_partial(\n",
    "    interaction_matrix_new,\n",
    "    user_features=user_features_matrix_new,\n",
    "    item_features=item_features_matrix_new,\n",
    "    epochs=best_epochs\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "identified-heaven",
   "metadata": {},
   "source": [
    "## Predict for a new sample user "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "liked-neutral",
   "metadata": {},
   "outputs": [],
   "source": [
    "def predict(model, dataset, user_item_pairs, user_features_matrix, item_features_matrix):\n",
    "    user_id_mapping, user_feature_mapping, item_id_mapping, item_feature_mapping = dataset.mapping()\n",
    "    predictions = model.predict(\n",
    "        user_ids=np.array([user_id_mapping[x] for x in user_item_pairs[\"user_id\"]]),\n",
    "        item_ids=np.array([item_id_mapping[x] for x in user_item_pairs[\"movie_id\"]]),\n",
    "        user_features=user_features_matrix,\n",
    "        item_features=item_features_matrix\n",
    "    )\n",
    "    \n",
    "    df_predictions = user_item_pairs.copy()\n",
    "    df_predictions[\"prediction\"] = predictions\n",
    "    \n",
    "    return df_predictions\n",
    "\n",
    "def merge_interactions(df_evaluation, interactions):\n",
    "    df_evaluation = df_evaluation.merge(interactions[[\"user_id\", \"movie_id\", \"rating\"]], on=[\"user_id\", \"movie_id\"], how=\"left\")\n",
    "    df_evaluation[\"has_interaction\"] = ~df_evaluation[\"rating\"].isna()\n",
    "    df_evaluation = df_evaluation.drop(columns=\"rating\")\n",
    "    return df_evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "white-luxury",
   "metadata": {},
   "outputs": [],
   "source": [
    "all_item_ids = np.unique(interactions[\"movie_id\"])\n",
    "\n",
    "user_sample_item_pairs = pd.DataFrame({\n",
    "    \"user_id\": new_users[2],\n",
    "    \"movie_id\": all_item_ids\n",
    "})\n",
    "\n",
    "predictions_sample = predict(model, dataset, user_sample_item_pairs, user_features_matrix, item_features_matrix)\n",
    "\n",
    "predictions_sample = merge_interactions(predictions_sample, interactions)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "shared-restaurant",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Condense item features into a compact list per item\n",
    "item_features_no_id = item_features.drop(columns=\"movie_id\")\n",
    "\n",
    "item_features_list_series = item_features_no_id.apply(\n",
    "    lambda row: [item_features_no_id.columns[i] for i, x in enumerate(row) if x == 1],\n",
    "    axis=1)\n",
    "\n",
    "item_features_list = pd.DataFrame({\n",
    "    \"movie_id\": item_features[\"movie_id\"],\n",
    "    \"feature_list\": item_features_list_series,\n",
    "})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "laden-efficiency",
   "metadata": {},
   "outputs": [],
   "source": [
    "predictions_sample = predictions_sample \\\n",
    "    .merge(movie_titles, on=\"movie_id\") \\\n",
    "    .merge(item_features_list, on=\"movie_id\")\n",
    "\n",
    "predictions_sample = predictions_sample.sort_values(\"prediction\", ascending=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "massive-arthritis",
   "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>user_id</th>\n",
       "      <th>movie_id</th>\n",
       "      <th>prediction</th>\n",
       "      <th>has_interaction</th>\n",
       "      <th>movie_title</th>\n",
       "      <th>feature_list</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>99</th>\n",
       "      <td>27</td>\n",
       "      <td>100</td>\n",
       "      <td>-183.367188</td>\n",
       "      <td>True</td>\n",
       "      <td>Fargo (1996)</td>\n",
       "      <td>[Crime, Drama, Thriller]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49</th>\n",
       "      <td>27</td>\n",
       "      <td>50</td>\n",
       "      <td>-183.502594</td>\n",
       "      <td>True</td>\n",
       "      <td>Star Wars (1977)</td>\n",
       "      <td>[Action, Adventure, Romance, Sci-Fi, War]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>287</th>\n",
       "      <td>27</td>\n",
       "      <td>288</td>\n",
       "      <td>-183.567810</td>\n",
       "      <td>True</td>\n",
       "      <td>Scream (1996)</td>\n",
       "      <td>[Horror, Thriller]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>27</td>\n",
       "      <td>9</td>\n",
       "      <td>-183.671555</td>\n",
       "      <td>True</td>\n",
       "      <td>Dead Man Walking (1995)</td>\n",
       "      <td>[Drama]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>120</th>\n",
       "      <td>27</td>\n",
       "      <td>121</td>\n",
       "      <td>-183.720093</td>\n",
       "      <td>True</td>\n",
       "      <td>Independence Day (ID4) (1996)</td>\n",
       "      <td>[Action, Sci-Fi, War]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>735</th>\n",
       "      <td>27</td>\n",
       "      <td>742</td>\n",
       "      <td>-183.813568</td>\n",
       "      <td>True</td>\n",
       "      <td>Ransom (1996)</td>\n",
       "      <td>[Drama, Thriller]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>502</th>\n",
       "      <td>27</td>\n",
       "      <td>508</td>\n",
       "      <td>-183.830414</td>\n",
       "      <td>True</td>\n",
       "      <td>People vs. Larry Flynt, The (1996)</td>\n",
       "      <td>[Drama]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>297</th>\n",
       "      <td>27</td>\n",
       "      <td>298</td>\n",
       "      <td>-184.066299</td>\n",
       "      <td>True</td>\n",
       "      <td>Face/Off (1997)</td>\n",
       "      <td>[Action, Sci-Fi, Thriller]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>285</th>\n",
       "      <td>27</td>\n",
       "      <td>286</td>\n",
       "      <td>-184.377060</td>\n",
       "      <td>True</td>\n",
       "      <td>English Patient, The (1996)</td>\n",
       "      <td>[Drama, Romance, War]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>245</th>\n",
       "      <td>27</td>\n",
       "      <td>246</td>\n",
       "      <td>-184.455063</td>\n",
       "      <td>True</td>\n",
       "      <td>Chasing Amy (1997)</td>\n",
       "      <td>[Drama, Romance]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>117</th>\n",
       "      <td>27</td>\n",
       "      <td>118</td>\n",
       "      <td>-184.460281</td>\n",
       "      <td>True</td>\n",
       "      <td>Twister (1996)</td>\n",
       "      <td>[Action, Adventure, Thriller]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>294</th>\n",
       "      <td>27</td>\n",
       "      <td>295</td>\n",
       "      <td>-184.529266</td>\n",
       "      <td>True</td>\n",
       "      <td>Breakdown (1997)</td>\n",
       "      <td>[Action, Thriller]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>368</th>\n",
       "      <td>27</td>\n",
       "      <td>370</td>\n",
       "      <td>-184.838440</td>\n",
       "      <td>True</td>\n",
       "      <td>Mary Reilly (1996)</td>\n",
       "      <td>[Drama, Thriller]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>280</th>\n",
       "      <td>27</td>\n",
       "      <td>281</td>\n",
       "      <td>-184.888855</td>\n",
       "      <td>True</td>\n",
       "      <td>River Wild, The (1994)</td>\n",
       "      <td>[Action, Thriller]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>147</th>\n",
       "      <td>27</td>\n",
       "      <td>148</td>\n",
       "      <td>-184.903641</td>\n",
       "      <td>True</td>\n",
       "      <td>Ghost and the Darkness, The (1996)</td>\n",
       "      <td>[Action, Adventure]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>912</th>\n",
       "      <td>27</td>\n",
       "      <td>925</td>\n",
       "      <td>-185.267471</td>\n",
       "      <td>True</td>\n",
       "      <td>Unforgettable (1996)</td>\n",
       "      <td>[Sci-Fi, Thriller]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>509</th>\n",
       "      <td>27</td>\n",
       "      <td>515</td>\n",
       "      <td>-185.285919</td>\n",
       "      <td>True</td>\n",
       "      <td>Boot, Das (1981)</td>\n",
       "      <td>[Action, Drama, War]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>122</th>\n",
       "      <td>27</td>\n",
       "      <td>123</td>\n",
       "      <td>-185.299347</td>\n",
       "      <td>True</td>\n",
       "      <td>Frighteners, The (1996)</td>\n",
       "      <td>[Comedy, Horror]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>243</th>\n",
       "      <td>27</td>\n",
       "      <td>244</td>\n",
       "      <td>-185.431885</td>\n",
       "      <td>True</td>\n",
       "      <td>Smilla's Sense of Snow (1997)</td>\n",
       "      <td>[Action, Drama, Thriller]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1004</th>\n",
       "      <td>27</td>\n",
       "      <td>1017</td>\n",
       "      <td>-185.945892</td>\n",
       "      <td>True</td>\n",
       "      <td>Trees Lounge (1996)</td>\n",
       "      <td>[Drama]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      user_id  movie_id  prediction  has_interaction  \\\n",
       "99         27       100 -183.367188             True   \n",
       "49         27        50 -183.502594             True   \n",
       "287        27       288 -183.567810             True   \n",
       "8          27         9 -183.671555             True   \n",
       "120        27       121 -183.720093             True   \n",
       "735        27       742 -183.813568             True   \n",
       "502        27       508 -183.830414             True   \n",
       "297        27       298 -184.066299             True   \n",
       "285        27       286 -184.377060             True   \n",
       "245        27       246 -184.455063             True   \n",
       "117        27       118 -184.460281             True   \n",
       "294        27       295 -184.529266             True   \n",
       "368        27       370 -184.838440             True   \n",
       "280        27       281 -184.888855             True   \n",
       "147        27       148 -184.903641             True   \n",
       "912        27       925 -185.267471             True   \n",
       "509        27       515 -185.285919             True   \n",
       "122        27       123 -185.299347             True   \n",
       "243        27       244 -185.431885             True   \n",
       "1004       27      1017 -185.945892             True   \n",
       "\n",
       "                             movie_title  \\\n",
       "99                          Fargo (1996)   \n",
       "49                      Star Wars (1977)   \n",
       "287                        Scream (1996)   \n",
       "8                Dead Man Walking (1995)   \n",
       "120        Independence Day (ID4) (1996)   \n",
       "735                        Ransom (1996)   \n",
       "502   People vs. Larry Flynt, The (1996)   \n",
       "297                      Face/Off (1997)   \n",
       "285          English Patient, The (1996)   \n",
       "245                   Chasing Amy (1997)   \n",
       "117                       Twister (1996)   \n",
       "294                     Breakdown (1997)   \n",
       "368                   Mary Reilly (1996)   \n",
       "280               River Wild, The (1994)   \n",
       "147   Ghost and the Darkness, The (1996)   \n",
       "912                 Unforgettable (1996)   \n",
       "509                     Boot, Das (1981)   \n",
       "122              Frighteners, The (1996)   \n",
       "243        Smilla's Sense of Snow (1997)   \n",
       "1004                 Trees Lounge (1996)   \n",
       "\n",
       "                                   feature_list  \n",
       "99                     [Crime, Drama, Thriller]  \n",
       "49    [Action, Adventure, Romance, Sci-Fi, War]  \n",
       "287                          [Horror, Thriller]  \n",
       "8                                       [Drama]  \n",
       "120                       [Action, Sci-Fi, War]  \n",
       "735                           [Drama, Thriller]  \n",
       "502                                     [Drama]  \n",
       "297                  [Action, Sci-Fi, Thriller]  \n",
       "285                       [Drama, Romance, War]  \n",
       "245                            [Drama, Romance]  \n",
       "117               [Action, Adventure, Thriller]  \n",
       "294                          [Action, Thriller]  \n",
       "368                           [Drama, Thriller]  \n",
       "280                          [Action, Thriller]  \n",
       "147                         [Action, Adventure]  \n",
       "912                          [Sci-Fi, Thriller]  \n",
       "509                        [Action, Drama, War]  \n",
       "122                            [Comedy, Horror]  \n",
       "243                   [Action, Drama, Thriller]  \n",
       "1004                                    [Drama]  "
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Movies that user rated\n",
    "predictions_sample[predictions_sample[\"has_interaction\"]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "id": "square-barcelona",
   "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>user_id</th>\n",
       "      <th>movie_id</th>\n",
       "      <th>prediction</th>\n",
       "      <th>has_interaction</th>\n",
       "      <th>movie_title</th>\n",
       "      <th>feature_list</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>257</th>\n",
       "      <td>27</td>\n",
       "      <td>258</td>\n",
       "      <td>-182.961700</td>\n",
       "      <td>False</td>\n",
       "      <td>Contact (1997)</td>\n",
       "      <td>[Drama, Sci-Fi]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>27</td>\n",
       "      <td>7</td>\n",
       "      <td>-183.247574</td>\n",
       "      <td>False</td>\n",
       "      <td>Twelve Monkeys (1995)</td>\n",
       "      <td>[Drama, Sci-Fi]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>221</th>\n",
       "      <td>27</td>\n",
       "      <td>222</td>\n",
       "      <td>-183.349976</td>\n",
       "      <td>False</td>\n",
       "      <td>Star Trek: First Contact (1996)</td>\n",
       "      <td>[Action, Adventure, Sci-Fi]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>312</th>\n",
       "      <td>27</td>\n",
       "      <td>313</td>\n",
       "      <td>-183.489288</td>\n",
       "      <td>False</td>\n",
       "      <td>Titanic (1997)</td>\n",
       "      <td>[Action, Drama, Romance]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>236</th>\n",
       "      <td>27</td>\n",
       "      <td>237</td>\n",
       "      <td>-183.688583</td>\n",
       "      <td>False</td>\n",
       "      <td>Jerry Maguire (1996)</td>\n",
       "      <td>[Drama, Romance]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>116</th>\n",
       "      <td>27</td>\n",
       "      <td>117</td>\n",
       "      <td>-183.699783</td>\n",
       "      <td>False</td>\n",
       "      <td>Rock, The (1996)</td>\n",
       "      <td>[Action, Adventure, Thriller]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>275</th>\n",
       "      <td>27</td>\n",
       "      <td>276</td>\n",
       "      <td>-183.765442</td>\n",
       "      <td>False</td>\n",
       "      <td>Leaving Las Vegas (1995)</td>\n",
       "      <td>[Drama, Romance]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>299</th>\n",
       "      <td>27</td>\n",
       "      <td>300</td>\n",
       "      <td>-183.785339</td>\n",
       "      <td>False</td>\n",
       "      <td>Air Force One (1997)</td>\n",
       "      <td>[Action, Thriller]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>256</th>\n",
       "      <td>27</td>\n",
       "      <td>257</td>\n",
       "      <td>-183.805420</td>\n",
       "      <td>False</td>\n",
       "      <td>Men in Black (1997)</td>\n",
       "      <td>[Action, Adventure, Comedy, Sci-Fi]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>281</th>\n",
       "      <td>27</td>\n",
       "      <td>282</td>\n",
       "      <td>-183.860794</td>\n",
       "      <td>False</td>\n",
       "      <td>Time to Kill, A (1996)</td>\n",
       "      <td>[Drama]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>180</th>\n",
       "      <td>27</td>\n",
       "      <td>181</td>\n",
       "      <td>-183.979492</td>\n",
       "      <td>False</td>\n",
       "      <td>Return of the Jedi (1983)</td>\n",
       "      <td>[Action, Adventure, Romance, Sci-Fi, War]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>126</th>\n",
       "      <td>27</td>\n",
       "      <td>127</td>\n",
       "      <td>-184.021484</td>\n",
       "      <td>False</td>\n",
       "      <td>Godfather, The (1972)</td>\n",
       "      <td>[Action, Crime, Drama]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>403</th>\n",
       "      <td>27</td>\n",
       "      <td>405</td>\n",
       "      <td>-184.061295</td>\n",
       "      <td>False</td>\n",
       "      <td>Mission: Impossible (1996)</td>\n",
       "      <td>[Action, Adventure, Mystery]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>585</th>\n",
       "      <td>27</td>\n",
       "      <td>591</td>\n",
       "      <td>-184.094696</td>\n",
       "      <td>False</td>\n",
       "      <td>Primal Fear (1996)</td>\n",
       "      <td>[Drama, Thriller]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>268</th>\n",
       "      <td>27</td>\n",
       "      <td>269</td>\n",
       "      <td>-184.190216</td>\n",
       "      <td>False</td>\n",
       "      <td>Full Monty, The (1997)</td>\n",
       "      <td>[Comedy]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>249</th>\n",
       "      <td>27</td>\n",
       "      <td>250</td>\n",
       "      <td>-184.195923</td>\n",
       "      <td>False</td>\n",
       "      <td>Fifth Element, The (1997)</td>\n",
       "      <td>[Action, Sci-Fi]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>331</th>\n",
       "      <td>27</td>\n",
       "      <td>333</td>\n",
       "      <td>-184.250458</td>\n",
       "      <td>False</td>\n",
       "      <td>Game, The (1997)</td>\n",
       "      <td>[Mystery, Thriller]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>110</th>\n",
       "      <td>27</td>\n",
       "      <td>111</td>\n",
       "      <td>-184.253235</td>\n",
       "      <td>False</td>\n",
       "      <td>Truth About Cats &amp; Dogs, The (1996)</td>\n",
       "      <td>[Comedy, Romance]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>465</th>\n",
       "      <td>27</td>\n",
       "      <td>471</td>\n",
       "      <td>-184.255936</td>\n",
       "      <td>False</td>\n",
       "      <td>Courage Under Fire (1996)</td>\n",
       "      <td>[Drama, War]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>621</th>\n",
       "      <td>27</td>\n",
       "      <td>628</td>\n",
       "      <td>-184.265732</td>\n",
       "      <td>False</td>\n",
       "      <td>Sleepers (1996)</td>\n",
       "      <td>[Crime, Drama]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     user_id  movie_id  prediction  has_interaction  \\\n",
       "257       27       258 -182.961700            False   \n",
       "6         27         7 -183.247574            False   \n",
       "221       27       222 -183.349976            False   \n",
       "312       27       313 -183.489288            False   \n",
       "236       27       237 -183.688583            False   \n",
       "116       27       117 -183.699783            False   \n",
       "275       27       276 -183.765442            False   \n",
       "299       27       300 -183.785339            False   \n",
       "256       27       257 -183.805420            False   \n",
       "281       27       282 -183.860794            False   \n",
       "180       27       181 -183.979492            False   \n",
       "126       27       127 -184.021484            False   \n",
       "403       27       405 -184.061295            False   \n",
       "585       27       591 -184.094696            False   \n",
       "268       27       269 -184.190216            False   \n",
       "249       27       250 -184.195923            False   \n",
       "331       27       333 -184.250458            False   \n",
       "110       27       111 -184.253235            False   \n",
       "465       27       471 -184.255936            False   \n",
       "621       27       628 -184.265732            False   \n",
       "\n",
       "                             movie_title  \\\n",
       "257                       Contact (1997)   \n",
       "6                  Twelve Monkeys (1995)   \n",
       "221      Star Trek: First Contact (1996)   \n",
       "312                       Titanic (1997)   \n",
       "236                 Jerry Maguire (1996)   \n",
       "116                     Rock, The (1996)   \n",
       "275             Leaving Las Vegas (1995)   \n",
       "299                 Air Force One (1997)   \n",
       "256                  Men in Black (1997)   \n",
       "281               Time to Kill, A (1996)   \n",
       "180            Return of the Jedi (1983)   \n",
       "126                Godfather, The (1972)   \n",
       "403           Mission: Impossible (1996)   \n",
       "585                   Primal Fear (1996)   \n",
       "268               Full Monty, The (1997)   \n",
       "249            Fifth Element, The (1997)   \n",
       "331                     Game, The (1997)   \n",
       "110  Truth About Cats & Dogs, The (1996)   \n",
       "465            Courage Under Fire (1996)   \n",
       "621                      Sleepers (1996)   \n",
       "\n",
       "                                  feature_list  \n",
       "257                            [Drama, Sci-Fi]  \n",
       "6                              [Drama, Sci-Fi]  \n",
       "221                [Action, Adventure, Sci-Fi]  \n",
       "312                   [Action, Drama, Romance]  \n",
       "236                           [Drama, Romance]  \n",
       "116              [Action, Adventure, Thriller]  \n",
       "275                           [Drama, Romance]  \n",
       "299                         [Action, Thriller]  \n",
       "256        [Action, Adventure, Comedy, Sci-Fi]  \n",
       "281                                    [Drama]  \n",
       "180  [Action, Adventure, Romance, Sci-Fi, War]  \n",
       "126                     [Action, Crime, Drama]  \n",
       "403               [Action, Adventure, Mystery]  \n",
       "585                          [Drama, Thriller]  \n",
       "268                                   [Comedy]  \n",
       "249                           [Action, Sci-Fi]  \n",
       "331                        [Mystery, Thriller]  \n",
       "110                          [Comedy, Romance]  \n",
       "465                               [Drama, War]  \n",
       "621                             [Crime, Drama]  "
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Top recommendations from the movies the user didn't watch\n",
    "predictions_sample[~predictions_sample[\"has_interaction\"]].head(20)"
   ]
  }
 ],
 "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": "3.9.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}