USING EXPLAINABLE RECOMMENDER SYSTEM FOR MOVIE RECOMMENDATION

Abstract

Recommender systems have become essential in many industries such as ecommerce, entertainment, and social media, where they help users discover relevant content and products. Common techniques include Collaborative Filtering (CF), Matrix Factorization (MF), and Content-based Filtering, each offering personalized recommendations based on user preferences and behavior. However, despite their success, these systems face significant challenges. One of the major limitations is the lack of transparency in the recommendation process. Traditional models, such as neighborhood-based CF and matrix factorization, often provide recommendations without offering clear explanations as to why certain items are suggested. This lack of interpretability can lead to a decrease in user trust, as users may not understand the reasoning behind the suggestions. Furthermore, some advanced methods, such as recent Matrix Factorization models, require additional data sources, such as item content, to generate explanations, which may not always be available. To address this limitation, Explainable Recommender Systems are gaining importance. By providing transparent and interpretable recommendations, XRS improves user trust and allows users to verify the validity of suggestions. In this thesis, we introduce a novel Explainable Matrix Factorization (EMF) technique that generates both accurate and explainable top N recommendations, using only rating data.