TEMUL: Tensor-Based Deep Learning Approach for Multi-Criteria Recommender System
Fatemeh Rakee, Ali Hamzeh, Niloofar Mozafari · IEEE Access · 2025
Recommender systems are essential components of modern information technology, particularly in e-commerce platforms, where they help users navigate information overload through personalized suggestions. Traditional models rely solely on overall ratings, neglecting the fact that users often evaluate items based on multiple criteria. This work proposes a multi-criteria recommender system using tensor factorization and deep learning techniques. Deep learning techniques has achieved tremendous success in many emerging fields such as natural language processing, image processing, and speech recognition. Lately, the use of deep learning in recommender systems have been frequently explored with inspiring results. However, the majority of these researches only consider single-criteria ratings in their predictions. Also, as far as we know, there is not yet any study which considers reliability concepts in deep multi-criteria recommender systems. So, our Tensor-based Deep learning approach for multi-criteria recommender system algorithm which abbreviated as TEMUL has three advantages. 1) It analyzes user preferences from various aspects based on multi-criteria ratings; 2) it proposes a new deep multi-criteria collaborative filtering method based on reliability. 3) Last but not least; the experiments on a real-world multi-criteria dataset show that the proposed algorithm prove to be very effective in terms of producing more accurate predictions and recommendations compared with the other state-of-the-art recommender systems.