An optimized deep learning-based aggregation for multi-criteria-based collaborative filtering

For the past few decades, collaborative filtering (CF) has been considered the most commonly used recommendation tool. It suggests items to its users based on their past and similar users’ preferences. Typically, recommendations have been generated using single-criteria or overall ratings (ORs). However, single-criteria-based recommendation systems cannot completely capture user interests, resulting in low recommendation accuracy. This work majorly concentrates on enhancing the recommendation accuracy by dealing with the drawbacks of single-criterion recommendation systems through the use of multi-criteria ratings (MCRs). And, it also aims to address common challenges in multi-criteria-based CF, like data sparsity and cold starts. Most of the literature on multi-criteria-based recommendation systems (MCRS) has addressed these problems using heuristic approaches. This work proposes an optimized deep neural network (ODNN) approach for multi-criteria-based recommendation systems, named MCRS-ODNN. The proposed model’s performance was assessed using both prediction and recommendation accuracy metrics, using the benchmark datasets like Movielens 1 M, Yahoo! Movies, TripAdvisor, and BeerAdvocate. The simulations showed that the proposed model outperformed the existing MCRS models on all the benchmark datasets.