2nd YearIntermediateData 3-4 weeks
Content-Based Movie Recommender
A recommender that suggests similar movies using TF-IDF over genres, cast, and plot keywords with cosine similarity, served through a small web UI.
Pythonpandasscikit-learnFlaskTMDB dataset
Find GitHub repos & source code →How to build it — step by step
- 1Data prep: Clean a movie metadata dataset and build a combined "tags" string per movie.
- 2Vectorise: Compute TF-IDF vectors and a cosine-similarity matrix between all movies.
- 3Recommend: For a chosen movie, return the top-N most similar titles.
- 4Serve: Expose a Flask endpoint and a simple search UI showing posters and scores.
Key features to implement
- ✓TF-IDF + cosine similarity
- ✓Top-N recommendations
- ✓Search by title
- ✓Poster and metadata display
- ✓Pre-computed similarity for speed
💡 Unique twist to stand out
Add a hybrid mode that blends content similarity with simple popularity weighting, and let users tune the blend.
🎓 What you'll learn
Feature engineering, TF-IDF, similarity metrics, and serving a model behind an API.