Extracting features from the MovieLens 100k dataset