Research On Group Recommendation Algorithm Based On Trust Network

To solve the puzzle and reduce the waste of time and resource produced by Information overload, the personalized recommendation appeared. Now most recommended services are oriented to individuals. But as the rapidly development of all kinds of the social network, the communication between people is becoming more and more convenient. When a group is in a face of a Selective problem, it needs to consider all group members’ requirement rather than someone’s requirements. So the traditional individual recommendation services can’t meet the requirements. It needs to study the group recommendation based on the individual recommendation.The way of group recommendation can be divided into two kinds. One way is aggregation model, which aggregate all the group members’preferences to forms the group’s preferences model, and then recommend to the group based on the model. The other way is aggregation recommendation. This way firstly recommends to the individuals, and then aggregates the member’s recommendations to get the group’s recommendations.All the two ways involves the problem of the aggregation strategy, and the study of the aggregation is the focus of academic research. This study analysis of the aggregation strategies through the in-depth research, and On the basis of analyzing the advantages and disadvantages existing aggregation strategy, then combines the trust network and the group recommendation. Through the analysis of the trust network structure, two new group recommendation algorithms are presented.The specific research works are as follows:(1)This study summarizes the research status at home and abroad about trust networks and groups recommendation, then Analysis the basis of the relevant theories and research. This study also Points out that the promotion of trust network for recommendation service function, and the influence of social factors on the group recommended. Takes the aggregation recommendation as the main aggregation strategy research object, than respectively analysis the collaborative filtering recommendation algorithm based on the user and on the users, so as to get individual score predicts, and provide aggregate object for the next step group recommended.(2) In the group, because of the differences of the influence among the group and the professional authority, different people have different influence to the group decision. So when recommend to a group, different people should have different weight. First, this study uses the PageRank algorithm and trust network to calculate the influence weight of users, and then uses the user’s score frequency to calculate weight of authority. Considering the two finally, group recommendation algorithm is proposed based on the weight of the users. The experimental analysis the influence of the group trust degree of saturation to RMSE. Through the comparison with other aggregation strategies, recommendation with interaction can effectively improve the recommendation quality.(3) Considering the reality of group decision-making is a process of negotiation communication, the negotiation communication between users is simulated in this paper, and three factors that influence the interaction are pointed out. The three factors are trust, personality and differences in preferences. On the basis of in-depth analysis of these three factors, group recommendation algorithm based on user interaction was proposed. Through the experiments on different aggregation strategies, it shows that the average strategy can achieve the best effect, and that recommendation after interaction can obtain better recommendation result.