Group Recommendation Based On Trust Metric

Group recommendation is a kind of special service type which can satisfy both each users' group member' common and personalized requirements and group's common demands.Trust is an important concept in social networks,which can influence users' decisions and help people decide how much to interact with others.Most of existing trust based group recommendation methods pay little attention to the diversity of trust sources,resulting in poor recommendation accuracy.In addition,the existing list based recommendation methods tend to ignore the extreme recommendation of some members,which may easily lead to one-sided recommendation results.To address the above problems,this thesis studies group recommendation and proposes a group recommendation method based on mixed trust measurement.The main work is as follows:1.To solve the problem that most existing trust based group recommendation methods pay little attention to the diversity of trust sources,this thesis proposes a hybrid trust measurement method(HTM)for group members.HTM firstly creates an attribute trust matrix as well as a social trust matrix based on user attributes and social relationships respectively.Secondly,HTM accomplishes a hybrid trust matrix based on the integration of these two matrices with the employment of Tanimoto coefficient.Finally,the trust threshold is set based on hybrid trust measurement results to set the trust weight of group members.Simulation results show that the hybrid trust measurement method for group members proposed in this thesis improves the accuracy of user classification(user trusted or untrusted)by using the trust relationship and behavior intention of group members.2.The mainstream recommendation method based on the top N items only considers the whole recommendation list and ignores the extreme recommendation,this thesis proposes the aggregation group recommendation list method based on the hybrid trust mechanism.This method firstly makes use of the existing user trust mechanism,and proposes to determine the list of recommended items in different groups from the perspective of local extension to the whole.Finally,an aggregate ranking method is proposed to generate the final group recommendation list ranking.This scheme can effectively improve the quality of group recommendation,improve the satisfaction of group users on the recommendation project,and more adapt to the personalized and generic needs of group users.3.Based on the above theories and methods,this thesis constructs a prototype system of group recommendation,and gives an application demonstration of movie recommendation.Prototype of the system development process to complete the demand analysis,general design,detailed design and implementation,and to achieve hybrid trust value measure,subgroups recommended list function modules,such as polymerization and groups recommended verify the feasibility of the proposed method and theory,to demonstrate the application shows the group based on trust measurement recommended in the practical application effect of scenarios.