P3 Recommendation Engine

Type a few movies you like, build a user-movie graph, then watch three-hop random walks turn shared taste into recommendations.

MovieLens small

Start by adding a MovieLens title, for example Toy Story (1995) or Matrix, The (1999).

Your liked movies

Use MovieLens titles. The year is optional when the title is unambiguous.

P3 walks from you to a liked movie, from that movie to one of the displayed users, and from that user to another liked movie.

Animation speed

1.25x

Movies on canvas

Users on canvas

Walks this run

1Your typed movie becomes an edge from You to that movie.
2The canvas first checks displayed users, then fills the rest with random MovieLens users who rated it at least 3.5 stars.
3Each P3 walk samples one of your movies, one displayed user linked to it, then one other movie that user liked.

Dataset: MovieLens ml-latest-small, generated 2018-09-26. User ids are anonymized. This demo uses ratings of 3.5 or higher as liked edges.

Cleared every time random walks start.

Count	Movie	Last path
Run random walks to build recommendations.