Current students

Recommender Systems (RS) are software tools used to provide suggestions for items or other entities to the users by exploiting various strategies. Neighborhood-based (NB) techniques are among the most common approaches used to provide recommendations. Despite their simplicity, thanks to their generalization capability, they can achieve state-of-the-art accuracy and even outperform more complex techniques (e.g., deep learning) in many scenarios. Moreover, their efficacy is not only related to the accuracy of recommendations, but they can enhance several aspects of the user experience of a recommender system.
In my research I combine NB models with different ML techniques, aiming at merging the best properties of the two types of approach. The results have shown interesting improvements in many different tasks, including:
;

;
• user-level confidence estimation highly correlated with recommendation accuracy;
;
• item cold-start recommendation with state-of-the-art accuracy;
;
• stability and accuracy enhancements of Matrix Factorization algorithms;
;
• bias reduction in offline evaluation of recommender systems.
;