Movie Recommendation — Collaborative Filtering

A simple item-based collaborative-filtering movie recommender built with pandas and the MovieLens dataset. Given a few movies a user has rated, it surfaces other movies that similar viewers also liked.

How it works

1. Load ratings.csv and movies.csv and merge on movieId.
2. Pivot into a user × movie ratings table.
3. Drop movies with fewer than 10 ratings (too noisy) and fill the rest with 0.
4. Compute a Pearson correlation matrix between every pair of movies.
5. Given a list of (movie_title, user_rating) tuples, weight each movie's correlation column by (rating - 2.5) — positive for liked, negative for disliked — and sum across all input movies to score candidates.

Result: the top-scoring movies are the recommendations.

Dataset

Uses the MovieLens small dataset (included in dataset/):

• movies.csv — movie IDs and titles
• ratings.csv — user ratings (0.5–5 stars)
• tags.csv, links.csv — unused by this script

Requirements

• Python 3
• pandas
• scipy

pip install pandas scipy

Running

python temp.py

Two demo profiles are baked in — a "romantic_lover" and an "action_lover" preference list — both print their top-20 recommended movies. Edit the lists at the bottom of temp.py to try your own.