Personalized Recommendation Fused With Trust-aware Network

In the era of information overload, the unique birth of recommendation system has built one bridge between information producers and information consumers. With the help of recommendation system, information consumers are able to select what they should be interested in from countless information, and in the meanwhile, information producers could make their information stand out from countless information. Most people always prefer to trust their friends in their daily life, and it is to say that the trust-aware network could simulate the real society prominently, and the social network could be used in the recommendation system. This paper studies popular recommendation systems firstly, and then proposes personalized recommendation based on explicit trust and implicit trust respectively.Firstly, this topic proposes the personalized recommendation combined with label propagation and trust diffusion. For the trust data is also faced with the problem of sparse data, the accuracy of the recommendation is unsatisfactory. The community discovery algorithm based on label propagation is proposed to discover the big community which belongs to each single user. According to the trust-aware network of each single user, the preprocess algorithm is proposed to predict the new trust relationship so as to extend the trust-aware network. The hybrid trust diffusion algorithm is proposed to make distinct difference in the trust degree between one single user and other users in the big community.Secondly, this topic proposes the personalized recommendation combined with user behavior and the change of interest. For the rating data is faced with the problem of sparse data and most datasets lack trust data, the accuracy of the recommendation is unsatisfactory. The recommendation algorithm based on the change of interest is proposed to get each user’s level of interest changes among different items. The recommendation algorithm of implicit trust based on user behavior is proposed to predict user’s potential implicit trust degree on items.The result of the experiment shows that the presented algorithm has distinct improvement in accuracy in comparison with the traditional recommendation algorithm.