PAENL: personalized attraction enhanced network learning for recommendation

Reviews and user–item interactions have been widely used to predict the behaviors of users. However, the sparsity of user–item interactions on datasets remains a major challenge in predicting user behavior. Most of the fusion user–item information and review information is predicted for user behavior prediction in a linear sense. However, this coarse-grained data fusion encounters difficulty in finding the complex relationship between different modal features. In this study, we propose a personalized attraction enhanced network learning for recommendation PAENL. The model consists of two modules: a user–item feature learning module and a review feature interaction module. In addition to the capability of modeling heterogeneity information by convolutional neural networks, PAENL can capture the essence of different users’ emotional reviews by the attention neural model in a nonlinear sense. Experiments are conducted on three real datasets and compared with a variety of mainstream advanced algorithms. The results demonstrate that the proposed algorithm PAENL significantly outperforms all state-of-the-art methods, and the attention mechanism can increase the interpretability of the user behavior prediction.