# 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.
# ==============================================================================
# Each user is responsible for checking the content of datasets and the
# applicable licenses and determining if suitable for the intended use.
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.