geekyroshan
Back to Projects
YouTube Recommendation System
Recommendation Systems

YouTube RecSys

Collaborative filtering with Matrix Factorization + BPR. Explains recommendations with 'Because you watched'.

About the Project

A collaborative filtering recommendation engine using Matrix Factorization with Bayesian Personalized Ranking (MF-BPR). The system learns user and item embeddings from interaction data and provides explainable recommendations by showing which previously watched items drove each suggestion. Trained on MovieLens-100K with a Streamlit web interface for exploration.

Key Features

  • Matrix Factorization with BPR (Bayesian Personalized Ranking) loss
  • 128-dimensional user and item embeddings
  • Explainable recommendations with 'Because you watched' feature
  • Adjustable popularity blending for discovery vs relevance
  • Interactive Streamlit UI with Recall@K evaluation

Impact

Demonstrates production-ready recommendation system with explainability - a key differentiator for user trust.

Tech Stack

PyTorchStreamlitBPR LossMovieLens-100K

Metrics

128-dim embeddings
Popularity blending
Recall@K evaluation

Interested in this project?

Let's discuss how I can build something similar for you.