Improved Transfer Learning Algorithm Based on Cross-domain in Recommendation System

In recommendation system, the single-domain recommendation algorithm faces many practical problems. The main collaborative filtering algorithms only use the information of the interaction between users and items, and can not avoid the problem of data sparsity in practical applications. This study focuses on the fusion of information from other fields to deal with sparse data, and solve the problem that it is difficult to find effective association from cross-fields. An associative feeling network model based on the deep neural network was proposed. It realizes transfer learning by mining the content information and information of items to obtain the optimal feature association, thus the accuracy of recommendation was optimized. The algorithm can deal with the problem of data sparsity in the recommendation system. Compared with some recently proposed algorithms, it shows an improved performance(15%~20%)and better expansibility. © 2020, Editorial Department, Journal of South China University of Technology. All right reserved.