A multitask recommendation algorithm based on DeepFM and Graph Convolutional Network
Liqiong Chen, Xiaoyu Bi, Guoqing Fan, Huaiying Sun · Concurrency and Computation Practice and Experience · 2022
Abstract For a long time, the problems of cold start and sparse data have always been the key problems to be solved by the recommendation system. Researchers usually use auxiliary information to deal with the aforementioned problems, thereby achieving the purpose of enhancing the recommendation effect. For example, the multitask feature learning framework (MKR) uses knowledge graphs as auxiliary information to enhance recommendations. However, the MKR algorithm has the problem of insufficient semantic information representation which affect the recommendation results. Thus, a multitask recommendation algorithm based on DeepFM and graph convolutional network (DeepFM_GCN) is proposed. The graph convolution network is used to deeply mine auxiliary entity information in the knowledge graph to supplement the sparse item semantics information in the recommendation task. Through the method of cross compression unit combined with Deep Neural Network to achieve feature sharing items and entities which to make up for the impact of insufficient feature representation. Then the DeepFM_GCN model utilizes DeepFM to deeply mine the interaction feature of users and items to avoid inaccurate items recommended to users. From the analysis of the experimental results, the DeepFM_GCN model can more fully explore user and item features, accordingly avoiding semantic ambiguity and improving prediction accuracy.