科研之友
微信
新浪微博
Facebook
分享链接
摘要
A theoretical framework to deal with incomplete preference information in social network group decision making (SN-GDM) is put forward. Firstly, the concept of the type-2 linguistic trust function is defined to model direct trust relationship of group experts in SN-GDM. It adopts the bilattice structure with of trust/distrust values that are all described by distributed linguistic function. Secondly, the type-2 uninorm trust propagation operator is investigated to connect the indirect trust relationship between group experts. The path-ordering induced order weighted averaging (P-IOWA) operator is proposed to aggregate to multiple trust propagation paths into a collective one. Then, the completed trust relationship between group experts is constructed to estimate the incomplete preference information in decision matrix. Combining the social network trust with the collaborative filtering, a comprehensive estimation method for incomplete information is proposed. In detail, the estimation method based the social network trust uses the actual reputation between experts as 'historic' actions. While the estimation method by the collaborative filtering utilizes the similar degree of decision as 'current' information. Then, the proposed comprehensive estimation method resembles the Bayesian approach, and consequently it can enhance the reliability for the evaluation of the incomplete elements in GDM. Hence, a consensus reaching process for GDM with incomplete preference information is proposed. Finally, an example is given to illustrate the use of the proposed method.
