WatchList is a hybrid group recommendation system for film and TV content using Letterboxd profile data and Netflix Watch History.
WatchList allows users to choose between the option of generating individual recommendations for themselves or generating group recommendations for groups up to 3 users. This research project seeks to develop WatchList as a film and television group recommendation system using a hybrid approach and machine learning techniques, that uses user data from OTT and/or reviewing platforms to generate a list of personalized recommendations, with the option to implement results filtering. WatchList adopts a hybrid approach, where content-based filtering and collaborative filtering are applied through Term Frequency Inverse Document Frequency, Cosine Similarity, and Singular Value Decomposition. For the group recommendation component, it aggregates the individual recommendations of all users and sorts it based on an expert-user-weighted approach.
- Choose between generating individual or group recommendations.
- For users that choose group recommendations, select the number of users (up to 3) involved.
- Insert any of the following data below:
- Netflix Watch History (in csv format)
- Letterboxd Username
- Optionally filter recommendation results based on attributes selected (year, genre, average rating).
- Generate a recommendation list of films and TV content based on user input provided.
- Generate individual or group profile analysis of the user(s) based on their input data
