Hybrid collaborative recommendation of co-embedded item attributes and graph features

In recent decades, personalized recommendation systems have attracted much attention from multiple disciplines for recommending interested products and services to users. Recommendation accentuates both the importance of feature learning tasks and the challenges posed by the sparsity of rating matrix. A common method for addressing the sparsity problem is to extend the feature space by the attributes of users and/or items. However, there are two main drawbacks in most existing recommendation methods. The first is the high computational cost of most existing recommendation models when using additional information from users and/or items to expand the feature space. The second problem is that it is difficult to obtain user additional information due to the high cost of acquiring tag knowledge and the increase in user privacy awareness. In this paper, we propose a novel and simple model to address the abovementioned issues, which employs a semi-autoencoder to co-embed the attributes and the graph features of the items for rating prediction (short for Item-Agrec). More specifially, a semi-autoencoder is introduced to learn the hidden nonlinear features of items for achieving a low computational cost, and thus the proposed Item-Agrec model can flexibly use side information from different sources. Meanwhile, in the case that it is not easy to obtain the user's additional information, we take the item's graph features and attributes into consideration for improving the accuracy of recommendation. Experiments on several real world datasets demonstrate the effectiveness of the proposed Item-Agrec compared with state-of-theart attribute-aware and content-aware methods. (c) 2021 Elsevier B.V. All rights reserved.