Research Of Deep Cross-network Recommendation Algorithm Based On Gated Recurrent Unit And Attention Mechanism

With the rapid development of information technology,people are faced with a serious problem of information overload.Traditional methods such as categories and keyword retrieval can no longer meet users' needs in many scenarios,so the recommendation algorithm that passively provides information for users has attracted wide attention from academia and industry.First of all,the research background and related basic concepts of the recommendation algorithm are introduced,and also introduces commonly used algorithms such as matrix factorization algorithms,attention mechanisms and recurrent neural networks,as well as commonly used loss functions and evaluation indicators.Then,a deep cross network model based on gated recurrent unit and attention mechanism is proposed.The model firstly uses the gated recurrent unit to extract features from the user history behavior,and then establishes the correlation between features and the model target through the attention mechanism.Finally,low-order and high-order cross features are automatically extracted by feature crossing layer,and the results are fused and output by Sigmoid function.Finally,a comparison experiment with the current mainstream recommendation algorithms on three public data sets to verify the effectiveness of the algorithm.Data set includes user history behavior,commodity data and other information.The experimental results show that the model has a certain improvement on the AUC index compared with other algorithms,indicating that the model can effectively mine the potential features of the data and has good practicability.