Trust Aware Recommendation using Deep Matrix Factorization Model

Abstract

Recommender Systems, a critical tool in the field of information filtering, have recently undergone extensive research and development in both academic institutions and business. But the majority of today's recommender systems struggle with the following issues: (1) The user-item matrix's huge scale and sparse data need an impression on efficiency of recommendations. Therefore, the majority of recommender systems struggle to deal with customers who have left minimal ratings. It is sometimes discussed to as a taciturn jump issue. (2) The orthodox recommender methods considered the independence and uniform distribution of all users. This presumption ignores any user connections, which is inconsistent with suggestions made in the real world. In order to more correctly and realistically represent recommender systems, we present a model with a new factor trust analysis that naturally takes into account the preferences of the users and their reliable friends. Therefore, Deep matrix factorization (DMF) technique incorporates both the unambiguous impact of reliable users on the forecast of items for an active user, building on top of a state-of-the-art recommendation algorithm, SVD++. The investigational outcomes exhibit that our approach outperforms cutting-edge skills in this context. Our research shows that modeling trust metrics significantly improves suggestion accuracy, Mean Absolute Error (MAE), Root Mean Square Error (RMSE) and  F-Measure parameter specially for users who are just starting out.