Oct 2024/AI/ML/course
Books Recommendation System
Readers struggle to discover relevant content when catalog and rating data are disconnected. I built this book recommendation system to anchor machine learning choices firmly in the context of user discovery. The goal was to prove that recommendation logic must serve a clear business framing rather than just being a math exercise.
I used a notebook-based workflow to experiment on the Kaggle Book Recommendation Dataset. The pipeline implements two distinct approaches. Content-based filtering reasons directly from item attributes. Collaborative filtering leverages the collective behavior of the user base. Both approaches are documented and evaluated side by side.
The biggest takeaway is that there is no single right way to recommend a book. Content-based logic solves the cold start problem for new items, while collaborative filtering captures nuanced community tastes. This project reinforced that recommendation systems succeed through thoughtful evaluation narratives rather than brute force modeling.