# Copyright 2021 NVIDIA Corporation. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
http://developer.download.nvidia.com/notebooks/dlsw-notebooks/merlin_hugectr_multi-modal-data-1-download-convert/nvidia_logo.png

MovieLens-25M: Download and Convert

The MovieLens-25M is a popular dataset in the recommender systems domain, containing 25M movie ratings for ~62,000 movies given by ~162,000 users.

In this notebook, we will download and convert this dataset to a suitable format for subsequent processing.

Getting Started

# External dependencies
import os
import time

import pandas as pd
from sklearn.model_selection import train_test_split

from nvtabular.utils import download_file

We define our base input directory, containing the data.

INPUT_DATA_DIR = "./data"

We will download and unzip the data.

from os.path import exists

if not  exists(os.path.join(INPUT_DATA_DIR, "ml-25m.zip")):
    download_file("http://files.grouplens.org/datasets/movielens/ml-25m.zip", 
              os.path.join(INPUT_DATA_DIR, "ml-25m.zip"))

Convert the dataset

First, we take a look on the movie metadata.

movies = pd.read_csv(os.path.join(INPUT_DATA_DIR, 'ml-25m/movies.csv'))
movies.head()
movieId title genres
0 1 Toy Story (1995) Adventure|Animation|Children|Comedy|Fantasy
1 2 Jumanji (1995) Adventure|Children|Fantasy
2 3 Grumpier Old Men (1995) Comedy|Romance
3 4 Waiting to Exhale (1995) Comedy|Drama|Romance
4 5 Father of the Bride Part II (1995) Comedy

We can see, that genres are a multi-hot categorical features with different number of genres per movie. Currently, genres is a String and we want split the String into a list of Strings. In addition, we drop the title.

movies = movies.drop(['title', 'genres'], axis=1)
movies.head()
movieId
0 1
1 2
2 3
3 4
4 5

We save movies genres in parquet format, so that they can be used by NVTabular in the next notebook.

movies.to_parquet(os.path.join(INPUT_DATA_DIR, "movies_converted.parquet"))

Splitting into train and validation dataset

We load the movie ratings.

ratings = pd.read_csv(os.path.join(INPUT_DATA_DIR, "ml-25m", "ratings.csv"))
ratings.head()
userId movieId rating timestamp
0 1 296 5.0 1147880044
1 1 306 3.5 1147868817
2 1 307 5.0 1147868828
3 1 665 5.0 1147878820
4 1 899 3.5 1147868510

We drop the timestamp column and split the ratings into training and test dataset. We use a simple random split.

ratings = ratings.drop('timestamp', axis=1)
train, valid = train_test_split(ratings, test_size=0.2, random_state=42)

We save the dataset to disk.

train.to_parquet(os.path.join(INPUT_DATA_DIR, "train.parquet"))
valid.to_parquet(os.path.join(INPUT_DATA_DIR, "valid.parquet"))

Next steps

If you wish to download the real enriched data for the movielens-25m dataset, including movie poster and movie synopsis, then proceed through notebooks 02-04.

If you wish to use synthetic multi-modal data, then proceed to notebook 05-Create-Feature-Store.ipynb, synthetic data section.