Research On Personal Recommendation Of E-commerce Based On On-line Social Network

In recent years,the rapid development of Internet,e-commerce keeps increasing trend,the growing amount of information,and the phenomenon of overload,so the electronic commerce faces great challenges in the process of operation.In order to solve this problem,the electronic commerce recommendation system arises at the historic moment.And with the development of online social network,in the electronic commerce registered users exist in many social relations,the social relations to recommended system with a lot of help,can improve the problems of the traditional recommendation algorithm,such as "thin"," cold start" and "scalability",etc.Based on personalized recommendation algorithm are studied,considering the trust relationship between users.Firstly the research background,research purpose and research status are summarized,the related research were summarized,the theory on the basis of the proposed in this paper,the research question.First,to the mining of social network,using the theory of the chain,use of social network users to represent the link between the frequency of trust between,discussed respectively under the deterministic information and uncertain information processing method,in view of the uncertainty information using the theory of uncertainty about the method obtained trust;Second of all,the social network users,the similarity between the mining main consideration specific label and user of the user's attention to score two aspects of the project,through the Pearson similarity method to calculate the similarity between the user;Third,users-project evaluation matrix,the excavation in the process of e-commerce personalized recommendation,interest degree is measured by a user behavior of product data in a very important parameter,but in practice,users-project evaluation data is often missing,so you need to put the missing data prediction,this paper,by using the improved matrix decomposition algorithm to forecast the missing data.Fourth,construction of personalized recommendation neural network algorithm,this paper with the current number of users on the similarity of the current project,the average similarity,trusted user number,average score credibility,user-matrix of the input variables,whether to buy by the current user to the current project for the output variable to the design of the algorithm.At last,an algorithm demonstrate the effectiveness of the proposed algorithm in this paper,based on 1048576 data experiment,the recommendation results of high precision,has certain progressive significance.