Linkboy

28 Jan 2023

Discovering new movies is exciting. Even better if the discovered movies are surprisingly good. I’ve been playing around with some ideas for generating better recommendations. By better I mean movies that are very different, yet share the same “hidden qualities” that will make you love them.

The theory behind the recommendations engine is explained here and the repo and how to run locally is available on Linkboy’s GitHub.

Search for a movie

To use the recommendation engine you’ll need to refer to movies by their corresponding movieId. This can be found using this search functionality. Simply search for the movie using the whole or a part of the title.

Now let’s do some magic!

Find paths

Taste is dynamic. What you like depends on your frame of reference. To fully appreciate a movie, it may be better to watch some other movies first. If you enter the movieIds of two movies below you’ll get a suggestion on movies to see (in order) to better appreciate the target movie.

Tune the algorithm to your personal taste

The recommendation engine operates on a generic taste profile. To generate better recommendations you can tune it using your own MovieLens ratings. If you’re using MovieLens, navigate to export your data and click ‘export ratings’. Then upload the resulting file to Linkboy using the file chooser below. You remain entirely anonymous and the only reference to your tuned recommendations is a randomly generated key.

---