Trust Network Analysis And The Application Of Recommendation In Social Commerce

Social commerce is a new model of engaging in e-commerce activities through social media,which is a combination of social media and e-commerce functions.The recommendation system based on socialized commerce is a key technology and tool for realizing the efficiency and effectiveness of socialized e-commerce.Due to the inherent virtuality and openness of the network itself,social network analysis,especially the trust relationship between social network entities and the subject's reputation management are important theoretical and applied research topics in recommender system of social commerce.This thesis focuses on the above issues,which starts with reviewing the research results of domestic and foreign scholars in recent years on building trustworthy networks and implementing business recommendations.Then the analysis of the most basic trust relationships is given to explore and reflect psychological changes in trust relationships,including the optimization of existing trust propagation and aggregation model.After that,the new evaluation system of reputation is built to have a more comprehensive reflection of the formation of network users' reputation,and then a reputation-based influence measurement model is developed.Finally,a variety of recommendation schemes are proposed considering trust and similarity in the context of social commerce which are verified with real data set Douban.Specifically,the main research contents and achievements of this paper are as follows:First,in terms of the analysis of trust relationships and the formation of trust networks,intuitionistic fuzzy theory is used to describe the level of trust among users,which reflects the ambiguity of trust and helps to improve the accuracy of trust descriptions and later measures.And according to the psychological mechanism of trust relationships,an improved trust propagation operator is proposed to make the trust propagation process more in line with actual situation of people.Based on factors such as trust propagation path length,path strength,and information reliability,multiple trust aggregation strategies are proposed and compared to show different value and scope of these strategies.Secondly,in the aspect of reputation calculation and influence analysis,with the aid of existing research and analysis and application scenarios,the factors affecting the formation of reputation are analyzed.Then considering about the existing calculation methods,the reputation calculation and evaluation system is reconstructed,in which reputation is divided into static reputation and dynamic reputation.And the corresponding model is provided to improve the comprehensiveness of the coverage of the reputation value.The thesis takes the reputation value into account and explores the calculation method of the influence to comprehensively assess the user's influence in the trust network.Finally,based on the different roles of trust and reputation in the recommendation of social commerce applications,three recommendation strategies that combine trust are proposed,including business recommendations based on trust relationships,recommendation methods based on reputation and influence,and hybrid recommendation programs which use the trust factor to optimize the traditional collaborative filtering method.Then the comparison is given to reflect the advantages and disadvantages of different recommended programs from the perspective of theoretical analysis.And based on the actual Douban network data set,calculations are developed to achieve the recommended process of different recommendations,which evaluates the rationality and application value to give more suggestions about social commerce.The results show that different methods have different influence on prediction and recommendation,in which trust prediction,combined recommendation and pipeline hybried recommandation are effective in dealing with data sparsity,and recommendation based on reputation has great advantages in cold start.