A matter of taste.
Give a few films a rating. Watch user-based collaborative filtering turn shared preferences into recommendations.
Your taste profile
Choose at least two ratings. Leave films you haven’t rated blank.
Your next watch
See the example viewer data
These are six fictional viewers and hand-authored ratings, not a production dataset.
People with similar ratings can offer useful clues.
For each fictional viewer, cosine similarity is calculated on films both of you have rated. Viewers need at least two overlapping ratings. Predictions are similarity-weighted averages of their scores for films you have not rated.
Inspired by my Headstarter AI collaborative filtering project. The small, synthetic dataset makes the method inspectable; the results are illustrative rather than a claim of recommendation accuracy.
Cosine similarity · collaborative filtering · weighted prediction