Research On Top-N Recommendation Algorithm Integrating Social Information

With the rapid development of internet and information technology,people have entered an era of information explosion.Enjoying the convenience of network life,users also face the problem of "information overload".The recommendation algorithm based on collaborative filtering is one of the effective solutions for the network platform to solve the information overload problem due to its simple and efficient principle.However,this kind of algorithm cannot get an accurate recommendation encountering data sparse and cold start problems.To solve these problems,scholars tried to introduce social information into matrix factorization,which effectively improved the recommendation performance.However,some social information also has the problem of sparse data,which leads to unsatisfactory recommendation effect.Therefore,it is necessary to consider fully mining social information to avoid the shortcomings of traditional recommendation algorithm.In this paper,the existing measurement methods of social trust and user influence are optimized and improved to make full use of social trust and user influence to improve recommendation performance.The research includes the following parts:(1)From the perspective of implicit trust,a trust measurement model based on user preference is proposed.After analyzing several shortcomings of the traditional user preference calculation method,the model uses the entity features of the knowledge graph as auxiliary information to supplement the original item features to mine the user's potential preferences for the item.Finally,a trust matrix is constructed based on feedback from user preferences combined with initial trust to alleviate the cold-starting problem in the user's historical rating data and to fully tap social trust.(2)Considering the related characteristics of social network topology,an improved structural hole algorithm MSH(Modified Structural Hole)is proposed to measure the influence of users in social networks.This algorithm uses a trust network as a social network,and obtains the neighbor user's information diffusion contribution ability based on its characteristics.The user influence is measured by combining the neighbor user's information diffusion contribution value and node importance.Experimental results on a variety of large-scale networks show that MSH achieves a faster propagation speed in the information dissemination model(SI),proving that MSH can effectively identify influential users in social networks.(3)Research on Top-N recommendation algorithms integrating social information.In order to make full use of social information to alleviate the shortcomings of traditional collaborative filtering,a social trust-based social algorithm FSTA(Factored similarity models for Social Trust Ameliorated)is proposed.This algorithm takes into account the differences in trusting relationships between users.From the perspective of trusters and trustees,they are integrated into the research of matrix factorization technology;in order to make full use of social information,we then integrated influence users on the basis of FSTA algorithm and proposed a socialized algorithm based on social trust and user influence FSTI(Factored similarity models with Social Trust and Influence)algorithm,and finally completed the Top-N recommendation ranking process.In order to verify the effectiveness of the two algorithms,extensive experiments were performed on two real-world datasets,Movie Lens-1M and Book-crossing.The experimental results prove that FSTA and FSTI have higher recommended performance than similar benchmark algorithms,of which FSTI performed the best.At the same time,cold-starting experiments show that the proposed two algorithms can maintain good performance even when the user-item interaction is sparse.