Session Recommend Based On Users' Long-short Term Interests And Social Networks

With the emergence of various mobile phone software and large-scale online shopping platforms,While these platforms provide convenience to users,they also lead to the problem of information overload.Therefore,recommendation systems are critical for displaying information of interest to users and increasing long-term user engagement.Session recommend that multiple items clicked or purchased by users over a period of time be handled as an event because of the time attribute,thus,session recommendations enable users to quickly capture their changing interests from their sequence of behaviors.So it has received widespread attention from academia and industry.Because the user's interests are influenced by the user's long-term habits,social networks,and the nature of the user and the project itself.As a result,there are deficiencies and challenges in session recommendation.Based on the existing research work,this paper focuses on the shortcomings of session recommendation model.By leveraging the user's long-term interests and the influence of social networks on user interests,while using the gating module to effectively integrate the user's long-term and short-term interests.This enables more accurate session recommendation.The main contributions of this paper is as follows.(1)Aiming at the problem of insufficient interactive extraction of user long-term interest features in session recommendation.Firstly,the attention mechanism and vector cross product operation are used to model various attributes of users and items so that users' long-term interests can fully interact.Then use the convolutional neural network to fully extract the longterm interest of the user.Combine it with the short-term interests of users into a unified recommendation model.Finally,the validity and reliability of the research on session recommendation in this paper are verified by using the existing public data sets.(2)This paper presents a model that combines social networks that have an impact on the user with the user's current session interests.The vector representation of users' friends is obtained by combining graph neural network and attention mechanism.Aiming at the problem of the integration of long-term/short-term interests of users in conversation recommendation.This paper designs a gated fusion method.Distribution of weights to users'long-term and shortterm interests.Make the fusion of long-term and short-term interests more effective.The accuracy of the whole model for predicting user interest is increased.Finally,the authenticity of our method is verified through experiment.