基于Louvain方法的社会网络大群体决策交互共识模型

An interactive consensus model for large scale group decision making with social network based on Louvain method

由于社会网络分析可以反映决策者在大群体决策(large group decision making,LGDM)共识问题中的社会关系,因此研究基于社会网络的LGDM共识问题具有现实意义。本文通过考虑决策者之间的社会网络关系,提出了一种交互式共识模型:首先,根据决策者之间的社会关系,采用Louvain方法对网络进行聚类,以降低大规模社会网络的复杂性;然后,基于可能性分布的犹豫模糊元素来表示每个子群的偏好,根据社会网络分析的中心性,计算决策者和子群权重;随后,给出三个层次的共识测量和反馈机制,来加速共识的达成过程。该模型在一致性调整过程中,整个社会网络也随之更新,允许子群分类的改变,而且考虑决策者的交互使结果更容易被接受。最后,通过案例分析验证方法的可行性,并通过与现有方法的比较,说明了方法的优越性。

Due to the growth of the large-scale interaction between users,some researchers have noticed the problem of Large Group Decision Making(LGDM).It is more difficult to obtain consensus among participants than traditional GDM issues.Therefore,reaching a consensus has become a basic issue for LGDM.As the LGDM consensus problems have begun to attract more attention,some researchers have solved the LGDM consensus problem by directly using the traditional GDM consensus model,but the result is with a large number of calculations and complex steps.Therefore,some researchers began to reduce the dimensions of LGDM problems through clustering algorithms.However,most traditional clustering algorithms used in these studies,such as K-means and PAM,require that k value is subjectively predetermined.Unfortunately,there are no specific theoretical guidelines to set the k value.In addition,researchers also proposed some models to improve the consensus of LGDM consensus problems.Those models are mainly divided into automatic methods and interactive methods.The automatic feedback mechanism is time-saving,while the interactive feedback mechanism requires expert interaction,but the results obtained are more accurate.However,both automatic and interactive methods have drawbacks.The results obtained by automated models are not precise enough,and interactive models require more time and effort.Moreover,most studies assume that all decision makers are independent,and few studies consider social relationships between decision makers.Social networks can not only enrich and improve the preference information of decision makers,but also promote the consensus process.Social networks have certain advantages in solving the complexity of LGDM issues.Some studies have done a great achievement to the application of SN in LGDM,but most of these studies only focus on the selection process and do not refer to consensus and consensus improving process.Therefore,by considering social network information among decision makers,a new interactive consensus model is proposed in this paper.First,according to the social relationships among decision makers,Louvain method is used to cluster the network to reduce the complexity of the large scale social network.Then hesitant fuzzy element based on possibility distribution is employed to represent the preference of each partition.According to the centrality of social networks analysis,weights for the DMs and the partitions can be calculated.Next,three levels of consensus measures and the feedback mechanisms are given to guide the consensus reaching process.The proposed model not only allows the partitions to change,but also takes personal feedback into account to make the decision more acceptable.Finally,an example is illustrated to demonstrate the feasibility of the proposed model for LGDM consensus problems,and the results are compared with the results of existing methods to verify the effectiveness of the proposed method.This paper studies the improved consensus model of LGDM based on social network analysis.The main conclusions include the following aspects:(1)Apply SNs to the LGDM consensus problem.Considering the influence of social network in subgroup classification and feedback mechanism,the formation of the consensus process can be accelerated,and the decision is more accurate.(2)In this paper,a large scale community detection algorithm,the Louvain method,is adopted to construct a social network of DMs to automatically cluster large scale networks without setting any parameters.The algorithm simple and efficient.(3)In the proposed feedback mechanism,we modify the DM’s FPRs through the interaction between DMs,and then update the entire social network.Afterwards,through the Louvain method,automatic clustering can be implemented to update the network and weight and complete the process of reaching consensus in the next round.The proposed feedback mechanism combines the advantages of the automatic mechanism and the interactive mechanism.It can not only accelerate the consensus formation process,but also make the decision making more accurate.