A Letterboxd recommender based on a user's previous ratings

Sam Learner is graphics journalist for the Financial Times and last year, he built a Letterboxd recommender that gathers movie ratings from any Letterboxd user and provide movie recommendations based on ratings data from thousands of other users.

A note on the methodology:

A user’s “star” ratings are scraped their Letterboxd profile and assigned numerical ratings from 1 to 10 (accounting for half stars). Their ratings are then combined with a sample of ratings from the top 4000 most active users on the site to create a collaborative filtering recommender model using singular value decomposition (SVD). All movies in the full dataset that the user has not rated are run through the model for predicted scores and the items with the top predicted scores are returned. Due to constraints in time and computing power, the maxiumum sample size that a user is allowed to select is 500,000 samples, though there are over five million ratings in the full dataset from the top 4000 Letterboxd users alone.

Like with any recommendation engine, the more data you have (in this case, Letterboxd ratings), the better recommendations you’ll get. You can also filter out well-known movies to give you more niche picks.

I tried it out with the sliders in the middle (so between faster and better results and all movies and less-reviewed movies) and my top 50 films were quite varied. Some I wouldn’t watch (Band of Brothers, Secrets & Lies, Steamboat Bill Jr.) and some I might (Richard Pryor: Live in Concert, Cowboy Bebop, and Hoop Dreams)

Letterboxd related: An Alternate Feminist Cinema list on Letterboxd

Leave a Reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.