Implemented a hybrid movie recommendation system that combines user interaction signals with content similarity from movie metadata.
Overview
The recommender uses collaborative filtering with ALS and content-based filtering with TF-IDF and cosine similarity. The system is designed to analyze user interactions and movie attributes to produce personalized recommendations.
Technical Focus
- ALS-based collaborative filtering
- TF-IDF vectorization of content metadata
- Cosine similarity for content-based ranking
- PySpark pipeline for scalable data processing
Repository
The full recommendation pipeline is available on GitHub.