Expert Opinion Aggregation And Evolutions In Group Decision Making

Group decision making is a process in which a panel of experts are invited to assess and rank a set of alternatives based on specific criteria and eventually form a group decision.It has been widely applied in various decision-making situations such as selection,personnel evaluation,etc.,as group decision-making shows its advantages in gathering the wisdom,backgrounds,knowledge of the group members.However,there are two challenges in group decision-making at present.One of the challenges is to reduce the individual’s subjectivity and imprecision to improve the group opinion’s reliability in the group decision-making.Another challenge is to capture the individuals’ behavior characteristics in the opinion formation process within an interpersonal network decision environment.These two challenges involve static opinion aggregation and opinion dynamics in group decision making,which has received wide attention recently.This thesis researches group decision-making problems from the static and dynamic perspectives.For the static group decision-making problem,we focus on the expert opinions aggregation problem with uncertain information.For the dynamic group decisionmaking problem,we investigate the opinion dynamics process influenced by social relations.The thesis consists of six chapters as follows.In Chapter 1,we introduce the backgrounds and fundamental concepts for the group decision-making problem.We also analyze the two challenges in group decision-making and then provide literature reviews of expert opinion aggregation and opinion dynamics.In Chapter 2,we investigate the expert opinion aggregation problem from the angle of optimization and explore how to aggregate the experts’ uncertain opinions to a group opinion.Then,we propose a new aggregation criterion that contains consensus level and confidence level with the experts’ opinions represented as probability density functions(PDFs).Subsequently,the expert opinion aggregation problem is formulated as a bi-objective optimization model to improve both objectivity and reliability in group decision making.The survey of professional forecasters(SPF)is used as an example to examine the feasibility and accuracy of the proposed approach.The result shows that the new expert opinion aggregation approach can improve the reliability of the group opinion.In Chapter 3,we further study the expert opinion aggregation problem from the angle of designing aggregation operators.First,we propose a new IOWA(induced ordered weighted averaging)operator with the experts’ opinions represented as continuous scalars,namely the average-induced OWA(AIOWA)operator.Subsequently,we extend the new operator to the situation where the experts’ opinions are represented as PDFs,namely AIOWA-PDF operator.The two aggregation operators facilitate capturing the distribution characteristics of the opinion concerning the consensus and construct a nonlinear aggregation of individual opinions.Last,we incorporate the constrained entropy-orness optimization model for determining weights into the proposed aggregation operators.Two case studies are conducted to show the effectiveness of the proposed operators.In Chapter 4,we explore the experts’ behaviors in a social network-based decision environment and study the dynamic opinion formation process.We capture the expert’s attitudinal characteristics on how the expert’s opinion is changed from the neighboring experts’ opinions.Then,we propose a new model for opinion dynamics in a social network based on the ordered weighted averaging(OWA)aggregation operator.A network structure-based convergence condition is given,and an analysis of the convergency of the model is also provided.Numerical illustrations with three typical topologies(circle,star,and fully connected,respectively)validate the effectiveness of the proposed model and indicate that experts’ attitudinal characters towards the neighboring experts’ opinions affect the consensus opinion and the convergence rate.In Chapter 5,we further investigate the experts’ opinion formation behaviors concerning the trust relationships among the experts.First,we define a new concept,namely ε-support degree,to explore the distribution characteristics of the neighboring experts’ opinions and present a trust relationship-based model for opinion dynamics.Subsequently,we present a new model for opinion dynamics inspired by the idea of the PID(proportion,integral,differential)controller,which captures the expert’s behaviors after analyzing the neighboring experts’ opinion distribution characteristics,the history modification behavior,and the opinion modification trends.Finally,we use an example containing 21 managers within a trust relationship network to verify the effectiveness of the two models for opinion dynamics.The result shows that the PID-inspired model within a trust relationship network can speed up the convergence rate of the opinion dynamics process.In Chapter 6,we conclude the thesis,point out the contributions and limitations of the research,and put forward the prospect of future research.