Recommender Systems

A deep dive into recommender systems using the MovieLens dataset, building and comparing multiple approaches from scratch rather than relying on prebuilt recommendation libraries.

After an in-depth exploration and cleaning of the dataset, a baseline predictor is implemented along with an evaluation framework based on 5-fold cross-validation, using RMSE, MAE, Precision@N, and Recall@N as metrics. On top of this baseline, the project implements and tunes user-based and item-based collaborative filtering, content-based filtering, matrix factorization (ALS), and a hybrid model, comparing their trade-offs between rating accuracy and ranking quality on a held-out test set.