MCRS-SAE: multi-criteria recommender system based on sparse autoencoder
Amir Reza Kalantarnezhad, Javad Hamidzadeh · 2022
Due to a large amount of information in today’s world, recommender systems help users reach the things that are most similar to their tastes. Compared to traditional recommender systems (RSs), Multi-criteria recommender systems, consider several minor criteria instead of a general criterion, which causes predictions to be more similar to the target user’s preferences. However, high dimension and sparsity of data are always one of the main problems that reduce the quality of predictions. In recent years, it has also been observed that the use of deep learning techniques in recommender systems has increased the quality of predictions. In addition, some criteria prioritize the user when selecting an item. In this paper, we propose a multi-criteria recommender system using the sparse autoencoder (MCRS_SAE) to improve the problem of the sparsity of data and also use the radial basis function kernel (RBF kernel) to determine the weight of the criteria. Experiments on the Yahoo! Movies and TripAdvisor multi-criteria datasets show that our proposed method of presenting personal predictions has a better performance than other presented methods.