source: https://grouplens.org/datasets/movielens/

GroupLens Research has collected and made available rating data sets from the MovieLens web site (http://movielens.org). The data sets were collected over various periods of time, depending on the size of the set.

This dataset (ml-20m) describes 5-star rating and free-text tagging activity from MovieLens, a movie recommendation service. It contains 20000263 ratings and 465564 tag applications across 27278 movies. These data were created by 138493 users between January 09, 1995 and March 31, 2015. This dataset was generated on March 31, 2015, and updated on October 17, 2016 to update links.csv and add genome-* files.

create virtual environment

$ python3.6 -m venv ./ENV

or

$ virtualenv --python=python3.6 ./ENV

enter virtual environment

$ source ./ENV/bin/activate

if you want to exit virual environment,

$ deactivate

install dependencies under virtual environment

$ pip3.6 install -r requirements.txt

$ make prepare

Splitting raw data into public and private dataset is according to the analysis result in ./prepare/ml-20m_analysis.ipynb.

Q1: Rating Problem: Design a system to predict an unknown rating when given userId and movieId (at ./src/data/test_q1.csv). Evaluate results by RMSE.

Q2: Ranking Problem: We defined that the rating > 3.0 as a favorite movie (at ./src/data/likes_pub.pkl). Design a system to recommend a top-10 favorite movies for a person, a movieId list which userId did not see before (provide userId at ./src/data/test_q2.txt). Evaluate results by MAP@10.

Q3: Content-based Problem: We defined that the rating > 3.0 as a favorite movie (at ./src/data/likes_pub.pkl) . Design a system to recommend a top-10 userId they may like a new movie (at ./src/data/test_q3.txt). We will give you some information of that new movie (at ./src/data/ref_movie_feature.pkl). Evaluate results by MAP@10.