Research On Bi-Level Diffusion Network Recommendation Algorithm Based On Trust Relationship

With the popularization of the Internet,the information age has gradually begun.The exponentially increasing data information makes it difficult for people to obtain the required information in a timely and accurate manner,which greatly reduces the utilization rate of information,that is,the "information overload" problem.To overcome this problem,researchers at home and abroad have proposed a representative information filtering technology-recommendation systems.It does not require users to provide clear needs to predict their interests and preferences,and effectively assists users to make accurate choices.With the rapid development of e-commerce,personalized recommendation technology has also achieved greater research breakthroughs.Today,recommendation systems have become a hot research area covering multidisciplinary knowledge.In recent years,research on complex networks and big data technologies has yielded many results.Therefore,research scholars have combined complex networks with recommendation systems,including bipartite graph-based recommendation algorithms,that is,the diffusion of research resources in a bipartite network.In order to further improve the performance of the recommendation algorithm,this paper has carried out a series of researches on improving the recommendation algorithm based on the diffusion process.The main research work includes the following points:(1)A trust relationship network is constructed based on user context information."People gather by category and things by group." Combined with the idea of collective wisdom,people with more similarities are more likely to become friends.Therefore,this paper draws on the principle of trust between people in social psychology,and generates a similarity-based trust relationship network from three aspects: user behavior information,user context information,and user interest preferences.It helps to alleviate the impact of data sparsity on the recommendation algorithm.(2)A hybrid recommendation algorithm based on material diffusion and heatconduction is proposed.Combining the user trust relationship network,this article expands the traditional “user-item” bipartite graph to a “user-user-item” two-layer network,so that the resource diffusion process takes place in the two-layer network.The traditional recommendation algorithm based on resource diffusion implements the resource diffusion according to the principle of the average distribution of node degrees.The algorithm proposed in this paper introduces a trust relationship so that resources are allocated according to the weight ratio when the user-user network layer spreads.At the same time,the introduction of user context information helps alleviate the problem of cold start of users.(3)The performance of several classic recommendation algorithms based on MovieLens dataset is compared.The proposed algorithms are compared with traditional collaborative filtering algorithms,recommendation algorithms based on substance diffusion,recommendation algorithms based on heat conduction,and hybrid recommendation algorithms of the former two.The experimental results show that the algorithm proposed in this paper can improve the accuracy and diversity of the recommendation to a certain extent,and help alleviate the long-tail effect.