Bookommender

Bookommender was a project I created together with 3 other students during my MMT master’s degree in the Recommender Systems lecture. The goal was to develop a book recommendation platform with the Book-Crossing Dataset as a basis.

For this purpose, we implemented a factorization machine for user-based and item-based recommendation. We evaluated our model to a most- and least-popular baseline in terms of precision, recall and the NDCG rank metric. In this project I was responsible for the containerized setup with Docker, the data cleansing and enhancement, which I implemented with Python, the search functionality, and some frontend tasks.

The book data was extended via the OpenLibrary API and the UI created with React. FastAPI was used for the backend and the RankFM library served as the basis for the recommendations. The search is done via the ElasticSearch engine in the background.