demos/OracleAutoMLx_Recommendation.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "b33f1c32",
   "metadata": {},
   "source": [
    "\n",
    "<h1><font color=red>Building a Recommender using AutoMLx</font></h1>\n",
    "<p style=\"margin-left:5%; margin-right:5%;\">by the <font color=teal> Oracle AutoMLx Team </font></p>\n",
    "\n",
    "***"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d4420aae",
   "metadata": {},
   "source": [
    "Recommendation Demo Notebook.\n",
    "\n",
    "Copyright © 2024, Oracle and/or its affiliates.\n",
    "\n",
    "Licensed under the Universal Permissive License v 1.0 as shown at https://oss.oracle.com/licenses/upl/"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3e132352",
   "metadata": {},
   "source": [
    "# Overview of this Notebook\n",
    "\n",
    "In this notebook we will build a recommender using the Oracle AutoMLx tool for the Movielens 100k dataset to predict the next item that users will most likely watch, based on their ratings history.\n",
    "We explore the various options provided by the Oracle AutoMLx tool, allowing the user to control the AutoMLx training process. We finally evaluate the different models trained by AutoMLx. Depending on the machine running this notebook, it can take up to minutes.\n",
    "\n",
    "---\n",
    "## Prerequisites:\n",
    "\n",
    "   - Experience level: Novice (Python and Machine Learning)\n",
    "   - Professional experience: Some industry experience\n",
    "---\n",
    "\n",
    "## Business Use:\n",
    "\n",
    "Data analytics and modeling problems using Machine Learning (ML) are becoming popular and often rely on data science expertise to build accurate ML models. Such modeling tasks primarily involve the following steps:\n",
    "- Preprocess dataset (clean, impute, engineer features, normalize).\n",
    "- Pick an appropriate model for the given dataset and prediction task at hand.\n",
    "- Tune the chosen model’s hyperparameters for the given dataset.\n",
    "\n",
    "All of these steps are significantly time consuming and heavily rely on data scientist expertise. Unfortunately, to make this problem harder, the best feature subset, model, and hyperparameter choice widely varies with the dataset and the prediction task. Hence, there is no one-size-fits-all solution to achieve reasonably good model performance. Using a simple Python API, AutoML can quickly jump-start the datascience process with an accurately-tuned model and appropriate features for a given prediction task.\n",
    "\n",
    "## Table of Contents\n",
    "\n",
    "- <a href='#setup'> Setup</a>\n",
    "- <a href='#load-data'> Load the Movielens 100k dataset</a>\n",
    "  - <a href='#coltypes'> Define the column types</a>\n",
    "  - <a href='#dataset-split'>Splitting the dataset</a>\n",
    "- <a href='#automl'> AutoML</a>\n",
    "  - <a href='#provider'> Create an Instance of AutoMLx</a>\n",
    "  - <a href='#default'> Train a Model using AutoMLx</a>\n",
    "  - <a href='#recommending'> Generate recommendations </a>\n",
    "  - <a href='#analyze'> Analyze the AutoMLx optimization process </a>\n",
    "      - <a href='#algorithm-selection'> Algorithm Selection</a>\n",
    "      - <a href='#hyperparameter-tuning'> Hyperparameter Tuning</a>\n",
    "  - <a href='#modellist'> Advanced AutoMLx Configuration</a>\n",
    "  - <a href='#custom-validation'> Use a custom validation set</a>\n",
    "  - <a href='#evaluation'> Final evaluation of the best model</a>\n",
    "\n",
    "<a id='setup'></a>\n",
    "## Setup\n",
    "\n",
    "Basic setup for the Notebook."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "615868b8",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-12-26T14:18:33.871934Z",
     "iopub.status.busy": "2024-12-26T14:18:33.871540Z",
     "iopub.status.idle": "2024-12-26T14:18:40.377293Z",
     "shell.execute_reply": "2024-12-26T14:18:40.376404Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:18:36,255] [automlx.backend] Overwriting ray session directory to /tmp/gbbsq6z5/ray, which will be deleted at engine shutdown. If you wish to retain ray logs, provide _temp_dir in ray_setup dict of engine_opts when initializing the AutoMLx engine.\n"
     ]
    }
   ],
   "source": [
    "\n",
    "\n",
    "import datetime\n",
    "import logging\n",
    "import os\n",
    "import time\n",
    "import urllib\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "from automlx import AutoRecommender, init\n",
    "\n",
    "# Settings for plots\n",
    "plt.rcParams[\"figure.figsize\"] = [10, 7]\n",
    "plt.rcParams[\"font.size\"] = 15\n",
    "\n",
    "# Silence unnecessary warnings\n",
    "logging.getLogger(\"sanerec.autotuning.parameter\").setLevel(logging.ERROR)\n",
    "\n",
    "# Initialize the parallelization engine of AutoMLx\n",
    "init(engine='ray', engine_opts={\"ray_setup\": {\"log_to_driver\": False}})"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6eec4c8a",
   "metadata": {},
   "source": [
    "<a id='load-data'></a>\n",
    "## Load Movielens 100k data\n",
    "Movielens 100k dataset is one of the most common public datasets for movie recommendation. It contains 100k ratings from about 1k users on 1.6k movies, some information about user demographic, and additional movie characteristics. For more information about this dataset, you can visit the [Movielens website](https://grouplens.org/datasets/movielens/100k/).\n",
    "\n",
    "In this demo, we use the ratings to train a movie recommendation model, exploiting AutoMLx to find the best recommendation model and hyperparameters to use in terms of recommendation accuracy.\n",
    "Therefore, we start retrieving and loading the ratings data of the Movielens 100k dataset.\n",
    "To make this notebook lighter and quicker, we also subsample the ratings in the dataset, keeping only 50%."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "fae5912e",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-12-26T14:18:40.380010Z",
     "iopub.status.busy": "2024-12-26T14:18:40.379607Z",
     "iopub.status.idle": "2024-12-26T14:18:43.293176Z",
     "shell.execute_reply": "2024-12-26T14:18:43.292310Z"
    },
    "lines_to_next_cell": 2
   },
   "outputs": [],
   "source": [
    "\n",
    "\n",
    "get_ipython().system(' wget https://files.grouplens.org/datasets/movielens/ml-100k/u.data --no-check-certificate -q -O ./ml100k_interactions.tsv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "c83d4b6f",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-12-26T14:18:43.295997Z",
     "iopub.status.busy": "2024-12-26T14:18:43.295319Z",
     "iopub.status.idle": "2024-12-26T14:18:43.341015Z",
     "shell.execute_reply": "2024-12-26T14:18:43.340404Z"
    }
   },
   "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>timestamp</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>43660</th>\n",
       "      <td>508</td>\n",
       "      <td>185</td>\n",
       "      <td>5</td>\n",
       "      <td>883777430</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>87278</th>\n",
       "      <td>518</td>\n",
       "      <td>742</td>\n",
       "      <td>5</td>\n",
       "      <td>876823804</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14317</th>\n",
       "      <td>178</td>\n",
       "      <td>28</td>\n",
       "      <td>5</td>\n",
       "      <td>882826806</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>81932</th>\n",
       "      <td>899</td>\n",
       "      <td>291</td>\n",
       "      <td>4</td>\n",
       "      <td>884122279</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>95321</th>\n",
       "      <td>115</td>\n",
       "      <td>117</td>\n",
       "      <td>4</td>\n",
       "      <td>881171009</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       user_id  movie_id  rating  timestamp\n",
       "43660      508       185       5  883777430\n",
       "87278      518       742       5  876823804\n",
       "14317      178        28       5  882826806\n",
       "81932      899       291       4  884122279\n",
       "95321      115       117       4  881171009"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "\n",
    "\n",
    "dataset = pd.read_csv(\n",
    "    \"./ml100k_interactions.tsv\",\n",
    "    sep=\"\\t\",\n",
    "    names=[\"user_id\", \"movie_id\", \"rating\", \"timestamp\"],\n",
    ").sample(frac=0.5, random_state=1)\n",
    "\n",
    "dataset.head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2b3dbea2",
   "metadata": {},
   "source": [
    "In order to be used for the recommendation task, the data must have a timestamp column that is used to infer the temporal order of the samples. We also require to set the timestamp column as index of the dataframes used in our AutoML pipelines.\n",
    "\n",
    "Movielens contains a `timestamp` column that contains the time when a rating was given, so we set it as index of our dataframe."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "ba70ef5d",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-12-26T14:18:43.343403Z",
     "iopub.status.busy": "2024-12-26T14:18:43.342866Z",
     "iopub.status.idle": "2024-12-26T14:18:43.350233Z",
     "shell.execute_reply": "2024-12-26T14:18:43.349772Z"
    }
   },
   "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",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>timestamp</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>883777430</th>\n",
       "      <td>508</td>\n",
       "      <td>185</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>876823804</th>\n",
       "      <td>518</td>\n",
       "      <td>742</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>882826806</th>\n",
       "      <td>178</td>\n",
       "      <td>28</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>884122279</th>\n",
       "      <td>899</td>\n",
       "      <td>291</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>881171009</th>\n",
       "      <td>115</td>\n",
       "      <td>117</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           user_id  movie_id  rating\n",
       "timestamp                           \n",
       "883777430      508       185       5\n",
       "876823804      518       742       5\n",
       "882826806      178        28       5\n",
       "884122279      899       291       4\n",
       "881171009      115       117       4"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "\n",
    "\n",
    "dataset = dataset.set_index(\"timestamp\")\n",
    "dataset.head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bb766ce0",
   "metadata": {},
   "source": [
    "<a id='coltypes'></a>\n",
    "### Define types of columns in the dataframe\n",
    "\n",
    "The recommendation task requires to define the two main entities involved in the recommendation:\n",
    "- the `recommendation`, which represents the entity type that is going to be recommended;\n",
    "- the `recommendation_subject`, which represents the entity type that receives the recommendation.\n",
    "\n",
    "For this reason, AutoML requires to indicate what are the columns in the dataset that refer to these two concepts, and, in particular, the two columns that contain their unique identifiers.\n",
    "\n",
    "In our demo we want to recommend movies (`recommendation`), identified by the `movie_id` column, to users (`recommendation_subject`), identified by the `user_id` column. We declare this binding in a python dictionary that we will reuse throughout the demo."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "20d51b70",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-12-26T14:18:43.352070Z",
     "iopub.status.busy": "2024-12-26T14:18:43.351674Z",
     "iopub.status.idle": "2024-12-26T14:18:43.354241Z",
     "shell.execute_reply": "2024-12-26T14:18:43.353797Z"
    }
   },
   "outputs": [],
   "source": [
    "\n",
    "\n",
    "col_types = {\"movie_id\": \"recommendation\", \"user_id\": \"recommendation_subject\"}"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8118c470",
   "metadata": {},
   "source": [
    "<a id='dataset-split'></a>\n",
    "## Splitting the dataset\n",
    "\n",
    "We split the dataset into training and test datasets using a leave-last-out technique.\n",
    "The training set will be used to create a Machine Learning model using Oracle AutoMLx, and the test set will be used to evaluate the model's performance on unseen data.\n",
    "\n",
    "The leave-last-out splitting technique consists in keeping in the test set only the last data sample, as determined by its timestamp, for each `recommendation_subject` (user in this case). All the other samples form the training set. This corresponds to the common next item recommendation use case, where given the history of all the past data concerning a `recommendation_subject` in the training set, we want to predict what should be recommended next to the same subject, and check if it corresponds to the actual sample in the test set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "4f801f67",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-12-26T14:18:43.355950Z",
     "iopub.status.busy": "2024-12-26T14:18:43.355575Z",
     "iopub.status.idle": "2024-12-26T14:18:48.754372Z",
     "shell.execute_reply": "2024-12-26T14:18:48.753623Z"
    }
   },
   "outputs": [],
   "source": [
    "\n",
    "\n",
    "training_data, test_data = AutoRecommender.train_test_split(data=dataset, col_types=col_types)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "de17c7cb",
   "metadata": {},
   "source": [
    "<a id='automl'></a>\n",
    "# AutoML\n",
    "\n",
    "<a id='provider'></a>\n",
    "## Create an instance of Oracle AutoMLx\n",
    "\n",
    "The Oracle AutoMLx solution provides a pipeline that automatically finds a tuned model given a prediction task and a training dataset. In particular, it allows finding a tuned model for any supervised prediction task, for example, classification or regression where the target can be binary, categorical or real-valued.\n",
    "\n",
    "In this demo we want a model that performs a recommendation task, so we create a pipeline of type `AutoRecommender`, and we configure it with default parameters. You can find the complete list of all the available parameters and their meaning in our documentation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "dd7c6313",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-12-26T14:18:48.757538Z",
     "iopub.status.busy": "2024-12-26T14:18:48.756569Z",
     "iopub.status.idle": "2024-12-26T14:18:48.760522Z",
     "shell.execute_reply": "2024-12-26T14:18:48.759949Z"
    }
   },
   "outputs": [],
   "source": [
    "\n",
    "\n",
    "automl_pipeline = AutoRecommender().configure()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2550f528",
   "metadata": {},
   "source": [
    "<a id='default'></a>\n",
    "## Train a model using AutoMLx\n",
    "\n",
    "The training data is passed to the `fit()` function which executes the model selection and hyperparameter tuning steps."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "611e8210",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-12-26T14:18:48.762605Z",
     "iopub.status.busy": "2024-12-26T14:18:48.762077Z",
     "iopub.status.idle": "2024-12-26T14:20:00.950534Z",
     "shell.execute_reply": "2024-12-26T14:20:00.949856Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:18:49,097] [automlx.interface] Dataset shape: (49055,3)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:18:49,166] [automlx.process] Running Model Generation\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:18:49,207] [automlx.process] Model Generation completed.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:18:49,250] [automlx.model_selection] Running Model Selection\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:19:39,708] [automlx.model_selection] Model Selection completed - Took 50.458 sec - Selected models: [['ItemKNNRecommender']]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:19:39,753] [automlx.trials] Running Model Tuning for ['ItemKNNRecommender']\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:19:59,182] [automlx.trials] Best parameters for ItemKNNRecommender: {'n_recommendations': 10, 'num_of_neighbors': 506, 'bias': 0.0001, 'hist_len': 10, 'reciprocal_ranking': False, 'normalize_scores': False, 'cache_users_states': True}\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:19:59,184] [automlx.trials] Model Tuning completed. Took: 19.431 secs\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:19:59,639] [automlx.interface] Re-fitting pipeline\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:19:59,646] [automlx.final_fit] Skipping updating parameter seed, already fixed by FinalFit_ea4e7717-4\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:20:00,886] [automlx.interface] AutoMLx completed.\n"
     ]
    }
   ],
   "source": [
    "\n",
    "\n",
    "automl_pipeline = automl_pipeline.fit(data=training_data, col_types=col_types)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "34708471",
   "metadata": {},
   "source": [
    "<a id='recommending'></a>\n",
    "## Generate recommendations\n",
    "\n",
    "Once the AutoML pipeline is completed, we predict 5 recommendations for a random user in the dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "b6513047",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-12-26T14:20:00.952936Z",
     "iopub.status.busy": "2024-12-26T14:20:00.952431Z",
     "iopub.status.idle": "2024-12-26T14:20:01.011187Z",
     "shell.execute_reply": "2024-12-26T14:20:01.010693Z"
    }
   },
   "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>score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>731</td>\n",
       "      <td>190</td>\n",
       "      <td>21.294812</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>731</td>\n",
       "      <td>96</td>\n",
       "      <td>20.401263</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>731</td>\n",
       "      <td>97</td>\n",
       "      <td>19.808079</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>731</td>\n",
       "      <td>181</td>\n",
       "      <td>19.745601</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>731</td>\n",
       "      <td>174</td>\n",
       "      <td>19.685528</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   user_id  movie_id      score\n",
       "0      731       190  21.294812\n",
       "1      731        96  20.401263\n",
       "2      731        97  19.808079\n",
       "3      731       181  19.745601\n",
       "4      731       174  19.685528"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "\n",
    "\n",
    "recommendation_subjects = test_data.sample(1)[['user_id']]\n",
    "automl_pipeline.predict(subjects=recommendation_subjects, n_recommendations=5)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fd477e94",
   "metadata": {},
   "source": [
    "<a id='analyze'></a>\n",
    "## Analyze the AutoMLx optimization process\n",
    "\n",
    "During the Oracle AutoMLx process for recommendation, a summary of the optimization process is logged, containing:\n",
    "- Information about the training data.\n",
    "- Information about the AutoMLx Pipeline, such as:\n",
    "    - Selected algorithm that was the best choice for this data;\n",
    "    - Selected hyperparameters for the selected algorithm.\n",
    "\n",
    "AutoMLx provides a `print_summary` API to output all the different trials performed."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "7291b01d",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-12-26T14:20:01.013145Z",
     "iopub.status.busy": "2024-12-26T14:20:01.012716Z",
     "iopub.status.idle": "2024-12-26T14:20:01.026771Z",
     "shell.execute_reply": "2024-12-26T14:20:01.026308Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<br><div style='text-align: center; font-weight: bold; font-size: 16px;'>General Summary</div>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>ManualSplit(Shuffle=False, Seed=7)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>SanerecMetric</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>ItemKNNRecommender</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>{'n_recommendations': 10, 'num_of_neighbors': 506, 'bias': 0.0001, 'hist_len': 10, 'reciprocal_ranking': False, 'normalize_scores': False, 'cache_users_states': True}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>25.1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3.9.21 (main, Dec 11 2024, 16:24:11) \\n[GCC 11.2.0]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<br><div style='text-align: center; font-weight: bold; font-size: 16px;'>Trials Summary</div>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th>Step</th>\n",
       "      <th># Samples</th>\n",
       "      <th># Features</th>\n",
       "      <th>Algorithm</th>\n",
       "      <th>Hyperparameters</th>\n",
       "      <th>Score (SanerecMetric)</th>\n",
       "      <th>All Metrics</th>\n",
       "      <th>Runtime (Seconds)</th>\n",
       "      <th>Memory Usage (GB)</th>\n",
       "      <th>Finished</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>Model Selection</td>\n",
       "      <td>48114</td>\n",
       "      <td>2</td>\n",
       "      <td>ItemKNNRecommender</td>\n",
       "      <td>{'n_recommendations': 10, 'num_of_neighbors': 100, 'bias': 25, 'hist_len': 20, 'reciprocal_ranking': False, 'normalize_scores': False, 'cache_users_states': True}</td>\n",
       "      <td>0.0882</td>\n",
       "      <td>{'hr': 0.08820403825717323}</td>\n",
       "      <td>1.0596</td>\n",
       "      <td>0.7370</td>\n",
       "      <td>Thu Dec 26 06:19:02 2024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>Model Selection</td>\n",
       "      <td>48114</td>\n",
       "      <td>2</td>\n",
       "      <td>AlsRecommender</td>\n",
       "      <td>{'n_recommendations': 10, 'iterations': 10, 'factors': 16, 'regularization': 0.01, 'cache_users_states': True}</td>\n",
       "      <td>0.0765</td>\n",
       "      <td>{'hr': 0.0765143464399575}</td>\n",
       "      <td>4.5466</td>\n",
       "      <td>0.6639</td>\n",
       "      <td>Thu Dec 26 06:19:01 2024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>Model Selection</td>\n",
       "      <td>48114</td>\n",
       "      <td>2</td>\n",
       "      <td>TRexxRecommender</td>\n",
       "      <td>{'n_recommendations': 10, 'embedding_dim': 32, 'sequence_length': 5, 'num_sampled': 100, 'dropout_rate': 0.2, 'num_blocks': 2, 'num_head': 4, 'l2_reg_embedding': 1e-06, 'dnn_activation': 'tanh', 'optimizer_name': 'lazyadam', 'optimizer_learning_rate': 0.01, 'future_blinding': False, 'embeddings_on_cpu': False, 'cache_users_states': False, 'negative_sampling_method': CandidateSamplingMethod.UNIFORM_CANDIDATE_SAMPLING, 'epochs': 10, 'batch_size': 512, 'verbose': 1, 'augment_data': True, 'early_stopping_patience': -1}</td>\n",
       "      <td>0.0733</td>\n",
       "      <td>{'hr': 0.07332624867162593}</td>\n",
       "      <td>36.8648</td>\n",
       "      <td>1.2107</td>\n",
       "      <td>Thu Dec 26 06:19:39 2024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>Model Selection</td>\n",
       "      <td>48114</td>\n",
       "      <td>2</td>\n",
       "      <td>BprRecommender</td>\n",
       "      <td>{'n_recommendations': 10, 'iterations': 10, 'factors': 16, 'regularization': 0.01, 'cache_users_states': True}</td>\n",
       "      <td>0.0446</td>\n",
       "      <td>{'hr': 0.044633368756641874}</td>\n",
       "      <td>0.2655</td>\n",
       "      <td>0.7351</td>\n",
       "      <td>Thu Dec 26 06:19:02 2024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>Model Tuning</td>\n",
       "      <td>48114</td>\n",
       "      <td>2</td>\n",
       "      <td>ItemKNNRecommender</td>\n",
       "      <td>{'n_recommendations': 10, 'num_of_neighbors': 505, 'bias': 0.0001, 'hist_len': 10, 'reciprocal_ranking': False, 'normalize_scores': False, 'cache_users_states': True}</td>\n",
       "      <td>0.0999</td>\n",
       "      <td>{'hr': 0.09989373007438895}</td>\n",
       "      <td>1.4546</td>\n",
       "      <td>0.6808</td>\n",
       "      <td>Thu Dec 26 06:19:55 2024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>Model Tuning</td>\n",
       "      <td>48114</td>\n",
       "      <td>2</td>\n",
       "      <td>ItemKNNRecommender</td>\n",
       "      <td>{'n_recommendations': 10, 'num_of_neighbors': 506, 'bias': 0.0001, 'hist_len': 10, 'reciprocal_ranking': False, 'normalize_scores': False, 'cache_users_states': True}</td>\n",
       "      <td>0.0999</td>\n",
       "      <td>{'hr': 0.09989373007438895}</td>\n",
       "      <td>1.1579</td>\n",
       "      <td>1.2015</td>\n",
       "      <td>Thu Dec 26 06:19:58 2024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>Model Tuning</td>\n",
       "      <td>48114</td>\n",
       "      <td>2</td>\n",
       "      <td>ItemKNNRecommender</td>\n",
       "      <td>{'n_recommendations': 10, 'num_of_neighbors': 506, 'bias': 0.0001, 'hist_len': 10, 'reciprocal_ranking': False, 'normalize_scores': False, 'cache_users_states': True}</td>\n",
       "      <td>0.0999</td>\n",
       "      <td>{'hr': 0.09989373007438895}</td>\n",
       "      <td>1.4794</td>\n",
       "      <td>1.2011</td>\n",
       "      <td>Thu Dec 26 06:19:55 2024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>Model Tuning</td>\n",
       "      <td>48114</td>\n",
       "      <td>2</td>\n",
       "      <td>ItemKNNRecommender</td>\n",
       "      <td>{'n_recommendations': 10, 'num_of_neighbors': 10, 'bias': 28.25660795027468, 'hist_len': 10, 'reciprocal_ranking': False, 'normalize_scores': False, 'cache_users_states': True}</td>\n",
       "      <td>0.0956</td>\n",
       "      <td>{'hr': 0.09564293304994687}</td>\n",
       "      <td>1.3933</td>\n",
       "      <td>1.1959</td>\n",
       "      <td>Thu Dec 26 06:19:52 2024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>Model Tuning</td>\n",
       "      <td>48114</td>\n",
       "      <td>2</td>\n",
       "      <td>ItemKNNRecommender</td>\n",
       "      <td>{'n_recommendations': 10, 'num_of_neighbors': 10, 'bias': 28.26160794927468, 'hist_len': 10, 'reciprocal_ranking': False, 'normalize_scores': False, 'cache_users_states': True}</td>\n",
       "      <td>0.0956</td>\n",
       "      <td>{'hr': 0.09564293304994687}</td>\n",
       "      <td>1.1074</td>\n",
       "      <td>0.6722</td>\n",
       "      <td>Thu Dec 26 06:19:55 2024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>Model Tuning</td>\n",
       "      <td>48114</td>\n",
       "      <td>2</td>\n",
       "      <td>ItemKNNRecommender</td>\n",
       "      <td>{'n_recommendations': 10, 'num_of_neighbors': 10, 'bias': 28.26160794927468, 'hist_len': 10, 'reciprocal_ranking': False, 'normalize_scores': False, 'cache_users_states': True}</td>\n",
       "      <td>0.0956</td>\n",
       "      <td>{'hr': 0.09564293304994687}</td>\n",
       "      <td>1.2934</td>\n",
       "      <td>0.6724</td>\n",
       "      <td>Thu Dec 26 06:19:54 2024</td>\n",
       "    </tr>\n",
       "    <tr>\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",
       "      <td>Model Tuning</td>\n",
       "      <td>48114</td>\n",
       "      <td>2</td>\n",
       "      <td>ItemKNNRecommender</td>\n",
       "      <td>{'n_recommendations': 10, 'num_of_neighbors': 10, 'bias': 0.0001, 'hist_len': 20, 'reciprocal_ranking': False, 'normalize_scores': False, 'cache_users_states': True}</td>\n",
       "      <td>0.084</td>\n",
       "      <td>{'hr': 0.08395324123273114}</td>\n",
       "      <td>1.5070</td>\n",
       "      <td>0.6691</td>\n",
       "      <td>Thu Dec 26 06:19:55 2024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>Model Tuning</td>\n",
       "      <td>48114</td>\n",
       "      <td>2</td>\n",
       "      <td>ItemKNNRecommender</td>\n",
       "      <td>{'n_recommendations': 10, 'num_of_neighbors': 10, 'bias': 0.0001, 'hist_len': 21, 'reciprocal_ranking': False, 'normalize_scores': False, 'cache_users_states': True}</td>\n",
       "      <td>0.084</td>\n",
       "      <td>{'hr': 0.08395324123273114}</td>\n",
       "      <td>1.2909</td>\n",
       "      <td>0.6714</td>\n",
       "      <td>Thu Dec 26 06:19:54 2024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>Model Tuning</td>\n",
       "      <td>48114</td>\n",
       "      <td>2</td>\n",
       "      <td>ItemKNNRecommender</td>\n",
       "      <td>{'n_recommendations': 10, 'num_of_neighbors': 505, 'bias': 25, 'hist_len': 20, 'reciprocal_ranking': False, 'normalize_scores': False, 'cache_users_states': True}</td>\n",
       "      <td>0.0829</td>\n",
       "      <td>{'hr': 0.08289054197662062}</td>\n",
       "      <td>1.3678</td>\n",
       "      <td>1.2010</td>\n",
       "      <td>Thu Dec 26 06:19:42 2024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>Model Tuning</td>\n",
       "      <td>48114</td>\n",
       "      <td>2</td>\n",
       "      <td>ItemKNNRecommender</td>\n",
       "      <td>{'n_recommendations': 10, 'num_of_neighbors': 506, 'bias': 25, 'hist_len': 20, 'reciprocal_ranking': False, 'normalize_scores': False, 'cache_users_states': True}</td>\n",
       "      <td>0.0829</td>\n",
       "      <td>{'hr': 0.08289054197662062}</td>\n",
       "      <td>1.3354</td>\n",
       "      <td>1.2011</td>\n",
       "      <td>Thu Dec 26 06:19:46 2024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>Model Tuning</td>\n",
       "      <td>48114</td>\n",
       "      <td>2</td>\n",
       "      <td>ItemKNNRecommender</td>\n",
       "      <td>{'n_recommendations': 10, 'num_of_neighbors': 752, 'bias': 25, 'hist_len': 20, 'reciprocal_ranking': False, 'normalize_scores': False, 'cache_users_states': True}</td>\n",
       "      <td>0.0818</td>\n",
       "      <td>{'hr': 0.0818278427205101}</td>\n",
       "      <td>1.2453</td>\n",
       "      <td>1.2065</td>\n",
       "      <td>Thu Dec 26 06:19:49 2024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>Model Tuning</td>\n",
       "      <td>48114</td>\n",
       "      <td>2</td>\n",
       "      <td>ItemKNNRecommender</td>\n",
       "      <td>{'n_recommendations': 10, 'num_of_neighbors': 753, 'bias': 25, 'hist_len': 20, 'reciprocal_ranking': False, 'normalize_scores': False, 'cache_users_states': True}</td>\n",
       "      <td>0.0818</td>\n",
       "      <td>{'hr': 0.0818278427205101}</td>\n",
       "      <td>1.2532</td>\n",
       "      <td>1.2065</td>\n",
       "      <td>Thu Dec 26 06:19:47 2024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>Model Tuning</td>\n",
       "      <td>48114</td>\n",
       "      <td>2</td>\n",
       "      <td>ItemKNNRecommender</td>\n",
       "      <td>{'n_recommendations': 10, 'num_of_neighbors': 10, 'bias': 0.0001, 'hist_len': 132, 'reciprocal_ranking': False, 'normalize_scores': False, 'cache_users_states': True}</td>\n",
       "      <td>0.0797</td>\n",
       "      <td>{'hr': 0.07970244420828905}</td>\n",
       "      <td>1.5203</td>\n",
       "      <td>0.6742</td>\n",
       "      <td>Thu Dec 26 06:19:55 2024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>Model Tuning</td>\n",
       "      <td>48114</td>\n",
       "      <td>2</td>\n",
       "      <td>ItemKNNRecommender</td>\n",
       "      <td>{'n_recommendations': 10, 'num_of_neighbors': 10, 'bias': 0.0001, 'hist_len': 133, 'reciprocal_ranking': False, 'normalize_scores': False, 'cache_users_states': True}</td>\n",
       "      <td>0.0797</td>\n",
       "      <td>{'hr': 0.07970244420828905}</td>\n",
       "      <td>1.5426</td>\n",
       "      <td>0.6726</td>\n",
       "      <td>Thu Dec 26 06:19:55 2024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>Model Tuning</td>\n",
       "      <td>48114</td>\n",
       "      <td>2</td>\n",
       "      <td>ItemKNNRecommender</td>\n",
       "      <td>{'n_recommendations': 10, 'num_of_neighbors': 10, 'bias': 0.0001, 'hist_len': 255, 'reciprocal_ranking': False, 'normalize_scores': False, 'cache_users_states': True}</td>\n",
       "      <td>0.0797</td>\n",
       "      <td>{'hr': 0.07970244420828905}</td>\n",
       "      <td>1.4628</td>\n",
       "      <td>0.6716</td>\n",
       "      <td>Thu Dec 26 06:19:55 2024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>Model Tuning</td>\n",
       "      <td>48114</td>\n",
       "      <td>2</td>\n",
       "      <td>ItemKNNRecommender</td>\n",
       "      <td>{'n_recommendations': 10, 'num_of_neighbors': 10, 'bias': 0.0001, 'hist_len': 256, 'reciprocal_ranking': False, 'normalize_scores': False, 'cache_users_states': True}</td>\n",
       "      <td>0.0797</td>\n",
       "      <td>{'hr': 0.07970244420828905}</td>\n",
       "      <td>1.2602</td>\n",
       "      <td>0.6713</td>\n",
       "      <td>Thu Dec 26 06:19:54 2024</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "\n",
    "\n",
    "automl_pipeline.print_summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b0f42742",
   "metadata": {},
   "source": [
    "We also provide the capability to visualize the results of each stage of the AutoMLx pipeline.\n",
    "\n",
    "<a id='algorithm-selection'></a>\n",
    "### Algorithm Selection\n",
    "\n",
    "The plot below shows the scores predicted by Algorithm Selection for each algorithm. The horizontal line shows the average score across all algorithms. Algorithms below the line are colored turquoise, whereas those with a score higher than the mean are colored teal. The selected algorithm is in orange."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "2c2c5e99",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-12-26T14:20:01.028571Z",
     "iopub.status.busy": "2024-12-26T14:20:01.028156Z",
     "iopub.status.idle": "2024-12-26T14:20:01.212411Z",
     "shell.execute_reply": "2024-12-26T14:20:01.211915Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x700 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "\n",
    "\n",
    "def plot_model_selection_scores(_pipeline):\n",
    "    # Each trial is a row in a dataframe that contains\n",
    "    # Algorithm, Number of Samples, Number of Features, Hyperparameters, Score, Runtime, Memory Usage, Step as features\n",
    "    trials = _pipeline.completed_trials_summary_[\n",
    "        _pipeline.completed_trials_summary_[\"Step\"].str.contains(\"Model Selection\")\n",
    "    ]\n",
    "    name_of_score_column = f\"Score ({_pipeline._inferred_score_metric[0].name})\"\n",
    "    trials.replace([np.inf, -np.inf], np.nan, inplace=True)\n",
    "    trials.dropna(subset=[name_of_score_column], inplace=True)\n",
    "    scores = trials[name_of_score_column].tolist()\n",
    "    models = trials[\"Algorithm\"].tolist()\n",
    "\n",
    "    y_margin = 0.10 * (max(scores) - min(scores))\n",
    "    s = pd.Series(scores, index=models).sort_values(ascending=False)\n",
    "\n",
    "    colors = []\n",
    "    for f in s.keys():\n",
    "        if f.strip() == _pipeline.selected_model_.strip():\n",
    "            colors.append(\"orange\")\n",
    "        elif s[f] >= s.mean():\n",
    "            colors.append(\"teal\")\n",
    "        else:\n",
    "            colors.append(\"turquoise\")\n",
    "\n",
    "    fig, ax = plt.subplots(1)\n",
    "    ax.set_title(\"Algorithm Selection Trials\")\n",
    "    ax.set_ylim(min(scores) - y_margin, max(scores) + y_margin)\n",
    "    ax.set_ylabel(\"Hit Rate\")\n",
    "    s.plot.bar(ax=ax, color=colors, edgecolor=\"black\")\n",
    "    ax.axhline(y=s.mean(), color=\"black\", linewidth=0.5)\n",
    "    plt.show()\n",
    "\n",
    "plot_model_selection_scores(automl_pipeline)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9ff55095",
   "metadata": {},
   "source": [
    "<a id='hyperparameter-tuning'></a>\n",
    "### Hyperparameter Tuning\n",
    "\n",
    "Hyperparameter Tuning is the last stage of the AutoMLx pipeline, and focuses on improving the chosen algorithm's score on the dataset. We use a novel iterative algorithm to search across many hyperparameter dimensions, and converge automatically when optimal hyperparameters are identified. Each trial represents a particular hyperparameter configuration for the selected model.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "2ece9b6b",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-12-26T14:20:01.214316Z",
     "iopub.status.busy": "2024-12-26T14:20:01.213859Z",
     "iopub.status.idle": "2024-12-26T14:20:01.402519Z",
     "shell.execute_reply": "2024-12-26T14:20:01.401986Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x700 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "\n",
    "\n",
    "def plot_hp_tuning_scores(_pipeline):\n",
    "    # Each trial is a row in a dataframe that contains\n",
    "    # Algorithm, Number of Samples, Number of Features, Hyperparameters, Score, Runtime, Memory Usage, Step as features\n",
    "    trials = _pipeline.completed_trials_summary_[\n",
    "        _pipeline.completed_trials_summary_[\"Step\"].str.contains(\"Model Tuning\")\n",
    "    ]\n",
    "    name_of_score_column = f\"Score ({_pipeline._inferred_score_metric[0].name})\"\n",
    "    trials.replace([np.inf, -np.inf], np.nan, inplace=True)\n",
    "    trials.dropna(subset=[name_of_score_column], inplace=True)\n",
    "    trials.drop(trials[trials[\"Finished\"] == -1].index, inplace=True)\n",
    "    trials[\"Finished\"] = trials[\"Finished\"].apply(\n",
    "        lambda x: time.mktime(datetime.datetime.strptime(x, \"%a %b %d %H:%M:%S %Y\").timetuple())\n",
    "    )\n",
    "    trials.sort_values(by=[\"Finished\"], ascending=True, inplace=True)\n",
    "    scores = trials[name_of_score_column].tolist()\n",
    "    score = []\n",
    "    score.append(scores[0])\n",
    "    for i in range(1, len(scores)):\n",
    "        if scores[i] >= score[i - 1]:\n",
    "            score.append(scores[i])\n",
    "        else:\n",
    "            score.append(score[i - 1])\n",
    "    y_margin = 0.10 * (max(score) - min(score))\n",
    "    fig, ax = plt.subplots(1)\n",
    "    ax.set_title(\"Hyperparameter Tuning Trials\")\n",
    "    ax.set_xlabel(\"Iteration $n$\")\n",
    "    ax.set_ylabel(\"Hit Rate\")\n",
    "    ax.grid(color=\"g\", linestyle=\"-\", linewidth=0.1)\n",
    "    ax.set_ylim(min(score) - y_margin, max(score) + y_margin)\n",
    "    ax.plot(range(1, len(trials) + 1), score, \"k:\", marker=\"s\", color=\"teal\", markersize=3)\n",
    "    plt.show()\n",
    "\n",
    "plot_hp_tuning_scores(automl_pipeline)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6f699aea",
   "metadata": {},
   "source": [
    "<a id='modellist'></a>\n",
    "## Advanced AutoMLx Configuration\n",
    "\n",
    "You can also configure the AutoRecommender pipeline with suitable parameters according to your needs."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "17ecd010",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-12-26T14:20:01.404418Z",
     "iopub.status.busy": "2024-12-26T14:20:01.403988Z",
     "iopub.status.idle": "2024-12-26T14:20:01.407480Z",
     "shell.execute_reply": "2024-12-26T14:20:01.407057Z"
    }
   },
   "outputs": [],
   "source": [
    "\n",
    "\n",
    "custom_pipeline = AutoRecommender().configure(\n",
    "    model_list=[  # Specify the models you want the AutoMLx to consider\n",
    "        \"ItemKNNRecommender\",\n",
    "        \"AlsRecommender\",\n",
    "        \"BprRecommender\",\n",
    "    ],\n",
    "    n_algos_tuned=2,  # Choose how many models to tune\n",
    "    search_space={  # You can specify the hyperparameters and ranges we search for each model\n",
    "        \"ItemKNNRecommender\": {\"num_of_neighbors\": {\"range\": [10, 30], \"type\": \"continuous\"}}\n",
    "    },\n",
    "    max_tuning_trials=20,  # The maximum number of tuning trials. Can be integer or Dict (max number for each model)\n",
    "    score_metric=\"recall\",  # Any of the metrics available, see the documentation for a list of supported values\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a8fa275e",
   "metadata": {},
   "source": [
    "<a id='custom-validation'></a>\n",
    "## Use a custom validation set\n",
    "\n",
    "You can specify a custom validation set that you want AutoMLx to use to evaluate the quality of models and configurations."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "9fa672c6",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-12-26T14:20:01.409283Z",
     "iopub.status.busy": "2024-12-26T14:20:01.408820Z",
     "iopub.status.idle": "2024-12-26T14:20:01.576448Z",
     "shell.execute_reply": "2024-12-26T14:20:01.575921Z"
    }
   },
   "outputs": [],
   "source": [
    "\n",
    "\n",
    "training_data, validation_data = AutoRecommender.train_test_split(data=training_data, col_types=col_types)\n",
    "\n",
    "\n",
    "# We run again the AutoML pipeline with the custom training/validation split we just created, and some advanced settings that we can specify directly in the fit method."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "122eaa38",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-12-26T14:20:01.578550Z",
     "iopub.status.busy": "2024-12-26T14:20:01.578116Z",
     "iopub.status.idle": "2024-12-26T14:20:11.647411Z",
     "shell.execute_reply": "2024-12-26T14:20:11.646791Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:20:01,738] [automlx.interface] Dataset shape: (49055,3)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:20:01,808] [automlx.process] Running Model Generation\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:20:01,853] [automlx.process] Model Generation completed.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:20:01,884] [automlx.model_selection] Running Model Selection\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:20:03,779] [automlx.model_selection] Model Selection completed - Took 1.895 sec - Selected models: [['ItemKNNRecommender', 'AlsRecommender']]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:20:03,892] [automlx.trials] Running Model Tuning for ['ItemKNNRecommender']\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:20:07,284] [automlx.trials] Best parameters for ItemKNNRecommender: {'n_recommendations': 10, 'num_of_neighbors': 10, 'bias': 0.010099998000000002, 'hist_len': 10, 'reciprocal_ranking': False, 'normalize_scores': False, 'cache_users_states': True}\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:20:07,286] [automlx.trials] Model Tuning completed. Took: 3.394 secs\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:20:07,439] [automlx.trials] Running Model Tuning for ['AlsRecommender']\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:20:09,915] [automlx.trials] Best parameters for AlsRecommender: {'n_recommendations': 10, 'iterations': 10, 'factors': 16, 'regularization': 0.01, 'cache_users_states': True}\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:20:09,916] [automlx.trials] Model Tuning completed. Took: 2.478 secs\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:20:10,336] [automlx.interface] Re-fitting pipeline\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:20:10,346] [automlx.final_fit] Skipping updating parameter seed, already fixed by FinalFit_b09b807d-a\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2024-12-26 06:20:11,503] [automlx.interface] AutoMLx completed.\n"
     ]
    }
   ],
   "source": [
    "\n",
    "\n",
    "custom_pipeline = custom_pipeline.fit(\n",
    "    training_data,\n",
    "    col_types,\n",
    "    validation_data,\n",
    "    time_budget=20,  # Specify time budget in seconds\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a4a116f9",
   "metadata": {},
   "source": [
    "Now that the custom AutoML pipeline is completed, we can generate recommendations.\n",
    "Note that the pipeline's `recommend` method is equivalent to `predict`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "d0570c10",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-12-26T14:20:11.649691Z",
     "iopub.status.busy": "2024-12-26T14:20:11.649353Z",
     "iopub.status.idle": "2024-12-26T14:20:11.709528Z",
     "shell.execute_reply": "2024-12-26T14:20:11.709017Z"
    }
   },
   "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>score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>731</td>\n",
       "      <td>97</td>\n",
       "      <td>19.207128</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>731</td>\n",
       "      <td>132</td>\n",
       "      <td>18.304316</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>731</td>\n",
       "      <td>176</td>\n",
       "      <td>18.152165</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>731</td>\n",
       "      <td>181</td>\n",
       "      <td>18.135894</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>731</td>\n",
       "      <td>96</td>\n",
       "      <td>17.946448</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   user_id  movie_id      score\n",
       "0      731        97  19.207128\n",
       "1      731       132  18.304316\n",
       "2      731       176  18.152165\n",
       "3      731       181  18.135894\n",
       "4      731        96  17.946448"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "\n",
    "\n",
    "custom_pipeline.recommend(subjects=recommendation_subjects, n_recommendations=5)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8e9d90cb",
   "metadata": {},
   "source": [
    "<a id='evaluation'></a>\n",
    "## Final evaluation of the best model\n",
    "\n",
    "Finally, we evaluate the best model found on the test data we have. If no metric is specified, the pipeline computes the score using the same metric used to run the Hyperparameter Tuning, which in this case is the Recall, as we defined at pipeline creation.\n",
    "\n",
    "In this example, instead, we ask the pipeline to perform the evaluation using Normalized Discounted Cumulative Gain (NDCG), a common ranking metric. Our online documentation provides the list of the available metrics and how they are computed."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "35cfa1b0",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-12-26T14:20:11.711426Z",
     "iopub.status.busy": "2024-12-26T14:20:11.711002Z",
     "iopub.status.idle": "2024-12-26T14:20:12.276522Z",
     "shell.execute_reply": "2024-12-26T14:20:12.275804Z"
    },
    "lines_to_next_cell": 2
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "16d45dfc37994da8a64e3bec306dc43c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/939 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "0.0300"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "\n",
    "\n",
    "get_ipython().run_line_magic('precision', '4')\n",
    "custom_pipeline.score(data=test_data, score_metric=\"ndcg\")"
   ]
  }
 ],
 "metadata": {
  "jupytext": {
   "formats": "ipynb,py:percent"
  },
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.21"
  },
  "widgets": {
   "application/vnd.jupyter.widget-state+json": {
    "state": {
     "00958ac2ce954dd59c341e0ba7e66dc3": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HTMLStyleModel",
      "state": {
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HTMLStyleModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "StyleView",
       "background": null,
       "description_width": "",
       "font_size": null,
       "text_color": null
      }
     },
     "09413596dc37422b92713ec2406f99a0": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "ProgressStyleModel",
      "state": {
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "ProgressStyleModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "StyleView",
       "bar_color": null,
       "description_width": ""
      }
     },
     "16d45dfc37994da8a64e3bec306dc43c": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HBoxModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HBoxModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "2.0.0",
       "_view_name": "HBoxView",
       "box_style": "",
       "children": [
        "IPY_MODEL_a3b9b60a977646df93250f1a99a7e101",
        "IPY_MODEL_5dc3d49831044d49b3b4ab611c2fbbb1",
        "IPY_MODEL_f624705138df45379402e86dfe609f75"
       ],
       "layout": "IPY_MODEL_54fea72fa52b436d89533cc50f5faf33",
       "tabbable": null,
       "tooltip": null
      }
     },
     "1f4f03ef7cef4af4b806095bc5ec1d51": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "2.0.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "2.0.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border_bottom": null,
       "border_left": null,
       "border_right": null,
       "border_top": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "31118d501b0947148cc6d5cb9fa7003e": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HTMLStyleModel",
      "state": {
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HTMLStyleModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "StyleView",
       "background": null,
       "description_width": "",
       "font_size": null,
       "text_color": null
      }
     },
     "54fea72fa52b436d89533cc50f5faf33": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "2.0.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "2.0.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border_bottom": null,
       "border_left": null,
       "border_right": null,
       "border_top": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "5dc3d49831044d49b3b4ab611c2fbbb1": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "FloatProgressModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "FloatProgressModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "2.0.0",
       "_view_name": "ProgressView",
       "bar_style": "success",
       "description": "",
       "description_allow_html": false,
       "layout": "IPY_MODEL_1f4f03ef7cef4af4b806095bc5ec1d51",
       "max": 939.0,
       "min": 0.0,
       "orientation": "horizontal",
       "style": "IPY_MODEL_09413596dc37422b92713ec2406f99a0",
       "tabbable": null,
       "tooltip": null,
       "value": 939.0
      }
     },
     "9928e37af3e746fba4d4631a42f97787": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "2.0.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "2.0.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border_bottom": null,
       "border_left": null,
       "border_right": null,
       "border_top": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "a3b9b60a977646df93250f1a99a7e101": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HTMLModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HTMLModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "2.0.0",
       "_view_name": "HTMLView",
       "description": "",
       "description_allow_html": false,
       "layout": "IPY_MODEL_e936f2f8167e46a8b417677f690fdb3e",
       "placeholder": "​",
       "style": "IPY_MODEL_00958ac2ce954dd59c341e0ba7e66dc3",
       "tabbable": null,
       "tooltip": null,
       "value": "100%"
      }
     },
     "e936f2f8167e46a8b417677f690fdb3e": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "2.0.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "2.0.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "2.0.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border_bottom": null,
       "border_left": null,
       "border_right": null,
       "border_top": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "f624705138df45379402e86dfe609f75": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "2.0.0",
      "model_name": "HTMLModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "2.0.0",
       "_model_name": "HTMLModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "2.0.0",
       "_view_name": "HTMLView",
       "description": "",
       "description_allow_html": false,
       "layout": "IPY_MODEL_9928e37af3e746fba4d4631a42f97787",
       "placeholder": "​",
       "style": "IPY_MODEL_31118d501b0947148cc6d5cb9fa7003e",
       "tabbable": null,
       "tooltip": null,
       "value": " 939/939 [00:00&lt;00:00, 1954.84it/s]"
      }
     }
    },
    "version_major": 2,
    "version_minor": 0
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}