An Interactive Fusion Recommendation Model based on Deep Neural Network
Meng Rao, Kaiming Wang, Yansong Liu, Chao Wang · 2025
Recommendation model can extract the information that the user interested from large-scale data by analyzing historical behaviors, which has become the most effective method to alleviate "information overload". However, the existing recommendation algorithms still face the problems of sparse data and cold start. Therefore, this paper proposes an interactive fusion recommendation model based on deep neural network—DIFR. The model combines the explicit rating and implicit rating data to extract the semantic meaning of the review and the score matrix decomposition to obtain the original features of the review and rating data. Then the user-score-item and user-review-item interaction subgraph are established respectively, and the feature representation of users and items is learned by convolution reinforcement of the attribute graph to capture their in-depth interaction. Further, multi-dimensional attention mechanism is used to improve the fusion ability of interactive features. And a recommended score is calculated finally. Experimental results fully demonstrate the effectiveness of this method.