Given a dataset of Crunchyroll user interactions, how would you identify patterns that could inform content recommendations?
