Enhancing Recommender Systems through Active Learning Strategies with Matrix Factorization

With the rise of information overload and a multitude of options, the significance of recommender systems as indispensable tools remains paramount. These systems offer personalized suggestions, greatly improving user experiences in navigating the vast array of available options. This study explores the application of Matrix Factorization combined with active learning techniques to enhance the accuracy of recommender systems and tackle frequent challenges like sparse data and the cold start issue. The active learning strategies employed encompass both personalized approaches, like similarity to profile, highest predicted, and binary predicted, as well as non-personalized methods including random, popularity, Gini, error, popgini, and poperror. By applying these strategies to Matrix Factorization using the MovieLens and GoodBooks datasets, the study demonstrates significant improvements over conventional approaches, highlighting the critical role of active learning in enhancing recommender systems to better capture diverse user preferences.