Recommender-Systems

This repo is about developing and configuring parameters of Recommender System for Book-Crossing Dataset. Predicting ratings for books according to input data. Two approaches were used for this task:

1) Matrix Factorization
2) Neural Network with Transformer Encoder layers

The best results of Transformer are:

- AP@K: k=9, AP=1.0;
        k=20, AP=0.83
- MSE: 1.228 (in 10-point system)
- RMSE: 2.62