基于信任传递的群体推荐可视化交互共识集成方法

Under the social network (SN) environment, multiple users usually participate in group activities according to their interest, and then they need group recommendation (GR). The key of GR is how to aggregate the individual user preferences into a group preference, and further reach a consensus, which need consider two aspects: (1) How to determine the trust relationship between group users;(2) How does the trust affect group interaction and consensus. To solve theses problems, this proposal will completely study the inner mechanism of the group consensus by combining the methods of trust propagation, visual interaction and information fusion operators. In particular, it relates threefold: (1) The method for modeling group trust (trust propagation and trust path aggregation methods, etc.); (2) Group interaction method affected by the trust relationship (trust based visual interaction/ group feedback/ feedforward/ absorption methods, etc.); (3) Trust based group consensus reaching method (group consistency via individual independence compatible model/conflict model, etc.). This proposal refines the research problem of social network group recommendation consensus aggregation (SN-GR-CA), and then establishes the trust based group visual interactive method (T-GVIM). It will not only enrich the connotation of e-commerce group recommendation, but also extend the coverage of group decision theory. Consequently, it has positive academic significance.

在社会网络环境下，多个用户根据共同的兴趣等参加群体活动，就需要进行群体推荐。群体推荐的关键是如何把个体用户的偏好聚集成群体偏好，并达成共识，其需要考虑：①如何确定群体用户之间的信任关系；②信任如何影响群体进行交互达成共识。针对群体推荐集成面临的新需求，本课题以信任传递、可视化交互和信息融合算子等为手段，深入研究群体达成共识的内在机理，具体包括：①群体用户信任建模方法（信任传递和信任路径聚合等）；②信任影响群体交互的作用机制（基于信任的可视化/群体反馈/前馈/吸纳方法等）；③基于信任的群体共识达成方法（群体一致性与个体独立性相容模型/冲突模型等）。本课题提炼出社会网络环境下群体推荐共识集成问题，并建立基于信任的群体可视化交互式方法，既丰富了电子商务群体推荐的内涵，也同时拓展了群决策理论的外延，具有积极的学术意义。

群体推荐的核心是群体如何达成共识，如何提高群体决策的效率是管理科学研究的前沿问题之一。本项目从社会网络信任的角度，建立了“基于信任的群体共识推荐方法”。首先对于群体决策交互过程所涉及的群体和个体决策行为，建立了群体一致性程度与个体和谐度测度。利用这两测度，分别建立了信任/非信任两类群体交互优化模型，较好地平衡了群体与个体的冲突性，并且能够以较小的成本达成群体满意一致性，揭示了信任对群体交互的作用机制。然后利用区间偏好信息对出了社会网络群体的信任关系建模，然后利用信任产生群体推荐意见，发现有利提高群体的共识，提出了基于社会网络信任和共识的群体推荐意见集结模型，从而提高了电子商务群体推荐系统的可靠性。