Research On The Recommendation Algorithm By Fusing Multiple Trust Networks

With the development of the Web technology and the sharp increase of the network data,the problem of information overload becomes prominent.The personalized recommendation system solves the problem of information overload very well.The collaborative filtering algorithm in the personalized recommendation system has been widely studied and applied,but it is still severely plagued by the data sparsity,the cold start problems and the vulnerability problem.With the popularity of social networking platform,the Internet has accumulated a lot of trust data.With the trust network introduced into the traditional collaborative filtering algorithm,it provides a possible idea to solve the problem of sparseness,cold start and the vulnerability.In this paper,the research status of recommendation algorithm is analyzed,and the trust network is introduced to improve the collaborative filtering algorithm.First,aiming at the shortcomings of the existing trust models,this paper integrates basic trust,reliability,influence and the four factors of ego orientation based on the McKinsey's trust formula,which enhances the comprehensiveness and rationality of the trust model.Secondly,based on the multi-trust model,this paper proposes a collaborative filtering algorithm that integrates multiple trust networks.The algorithm generates the nearest neighbor through comprehensive consideration of the trust and the similarity.The experimental result shows that the algorithm is optimized in accuracy and diversity,and it also solves the problem of cold start-up and vulnerability,but the algorithm is still affected by data sparsity.Finally,by using the transferability and implicitly of trust to expand and fill the multi-trust network and raise the improved algorithm of MT-CF based on the trust extension.The experimental result shows that the improved algorithm achieves obvious optimization in prediction accuracy and diversity.It shows the effectiveness of the expansion to the improved algorithm on data,thus to alleviate the data sparseness problem.In addition,the ability to solve the problem of cold start-up and vulnerability is further enhanced.