Year 2025

Hybrid e-commerce recommender

A production-style recommender blending ALS collaborative filtering with content similarity, MMR re-ranking for diversity and a popularity fallback for cold starts, with one-click offline evaluation.

Highlights

• ALS implicit matrix factorisation plus TF-IDF/KNN content similarity
• Recall@K, MAP@K, NDCG@K, coverage, diversity and novelty metrics
• Streamlit UI with light and dark themes

Stack

Python · Streamlit · scikit-learn · implicit · pandas