Study On Group Decision Methods Based On Particle Swarm Optimization

In order to deal with a complex group decision making problem,a group of experts are commonly invited to discuss and express their opinions until a final solution accepted by all people is reached.During this process,one of the most important issues is that the judgment matrix expressed by a decision maker should have some internal logicality.It can not be self-contradictory,or in other words,it should be consistent in some degree.For the purpose of reaching consensus among the members of group,experts need to interactively discuss and reconcile.Therefore,it requires a certain level of flexibility.Information granularity is viewed as an essential conception asset,which provides a decision maker a tangible level of flexibility to adjust his/her initial position.The PSO algorithm can be used to model the interactive process of forming decisions in the group.This thesis studies the group decision making methods based on the PSO method,the main results are given as follows:(1)Some shortcomings of the given consensus model in an existing literature are pointed out.A modified group decision model is proposed and solved by the novel PSO algorithm.As.compared with the original paper,it is found that the new group decision model can effectively avoid the deficiencies of the model in the original one,and it implies the feasibility of the presented method.(2)A group decision making model is constructed by using multiplicative reciprocal matrices with flexibility.The objective function is established from two point of view.One is according to the deviation degree between each multiplicative reciprocal matrix and the collective matrix.The other is the consistency in the whole group.It is further optimized by the PSO algorithm to give a collective matrix.The effects of the parameters on the ranking of alternatives are analyzed.The comparisons with the existing models show the applicability of the proposed one.(3)A group decision model is established in virtue of additive reciprocal matrices under a certain flexibility.Inspired by the TOPSIS method,an ideal additive consistent matrix is formed according to individual ones.The objective function is built by considering two factors.The first is the deviation degree between each additive reciprocal matrix and the ideal one.The second is the consistency of the whole group.It is further optimized by the PSO algorithm and some comparison and discussion are given.The studies reveal that the particle swarm optimization offers a new way to solve the practical group decision problems and enriches the decision-making methodology.