A self-contained tool for exploring item–item similarity. Look up the nearest neighbours of any title, or rate a handful of films and get recommendations scored by similarity-weighted ratings.
Load your data
Everything runs in your browser — nothing is uploaded anywhere.
Movie list
movies.csv · movieid, title
Similarity matrix
movie_sim.csv · ~59 MB, Pearson
Movie
Show top
Add a movie you've seen
Recommend top
How recommendations are scored
Each movie you rate contributes only its 10 most similar films. Each candidate
movie j you haven't rated gets a recommendation score — the similarity-weighted sum
of your ratings, over the rated movies i whose top-10 it falls into:
score(j) = Σi sim(i,j)·ri
This is a relevance score, not a predicted star rating: it isn't divided by the total similarity, so
a film that's strongly similar to a movie you rated highly — or close to several of them —
ranks above one that barely made a top-10 list. Higher means a stronger recommendation. Movies you've
already rated are excluded.