Mechanism Research Of Sub-group Evaluation

In many comprehensive evaluation activities,affected by the complexity of evaluation objects,fuzziness of information,incomplete and biased judgments,we often need to build a group made up of multiple evaluation subjects and aggregate their opinions in specialized operator to get the final conclusion,which is known as "group evaluation".Furthermore,when the group size and evaluation objects becomes more and more,or the evaluation subjects and objects are too complex,even with significant heterogeneity,to avoid opinions from different subgroup become incomparable and unable to integrated,we usually requires grouping method to improve the efficiency and to save the cost,this paper defines this technology as "sub-group evaluation".Although there is a lot of evaluation organization form based on subgroup,no matter researchers from the community of comprehensive evaluation,group decision making or system science,no one has made a thorough and systematic study on this field so far,especially about how to design evaluation mechanisms according to different situations in practice.Based on the framework of subgroup evaluation,this paper uses consensus measurement technology,combined with statistical sampling and grouping theory,discusses the general problems of integration and construction of evaluation mechanism,so as to provide reference for the future study,the main work summarized as follows:1.Based on the homogeneity assumption,this paper expands evaluation methods into subgroup dimension,and puts forward a simple standardized correction mechanism,a mechanism with information feedback and a mechanism consider subject weight.Firstly,considering the differences of value mainly influenced by experts' evaluation scale and cognition,which performances as the differences on mean and standard,we put forward the simple standardized correction mechanism.Secondly,for effective integrate experts' evaluation information,and improving the acceptability of evaluation conclusion,we propose a mechanism based on information feedback.The main difference with the simple standardized correction mechanism is that it uses interactive method to reach consensus,that is,when opinions between experts did not pass consensus degree after inspection,experts should modify their opinions according to the information feedback from the organizers.Finally,we construct a mechanism considering the change of subjective weigh by taking into account of individual evaluators will give inconsistent opinions for cognition,experience or other subjective factors,either influencing evaluation efficiency by refusing to modify their opinions.This method can effectively reduce the influence of inconsistent opinion to final conclusion by giving the low consensus degree experts less weigh.2.We expand the subgroup evaluation techniques and design an evaluation mechanism based on equilibrium theory and a self-organization mechanism by considering the existence of leaders respectively.Among them,in order to ensure the later evaluation in line with the homogeneity assumption,the former mechanism achieves balanced grouping according to the first evaluation conclusion,and then using simple standardized correction mechanism to implement comparability between different subgroups;the latter mainly against the problem of subjective and objective constraints and excessive authority of leaders in group evaluation.Through selective evaluation to determine the reliability and validity of experts and using self-regression function to correct opinions,experts' opinions can reach consistency to leaders'opinion,which effectively avoiding the impact of time limit or opinions reluctant modified by experts to evaluation activity.3.In view of the subject and object in the realistic evaluation activity more accord with heterogeneity hypothesis,this paper puts forward the idea of "chain" group evaluation.The traditional group evaluation assumes evaluation subjects and objects are homogeneous,that is,the different evaluation objects within the subgroups are approximate in distribution,mean and variance on the whole.Based on object in real evaluation activities more in line with the characteristics of heterogeneity,we propose to construct "public factor" or "common evaluation factor" to act as the "bridge" or "media" to converse values between different groups,which can realize the original scores being equivalence between different subgroups and defines as "chain subgroup evaluation".4.Based on the heterogeneous subgroup evaluation framework,we propose a chain evaluation mechanism based on information entropy,a two-stage subgroup evaluation mechanism and a dynamic group evaluation method considers subjects vary.(1)In the face of aggregation issues of subgroup in the process of group evaluation,this paper proposes a chain evaluation mechanism based on information entropy theory.This method filters evaluation information according to the theory of information entropy at first,then uses evaluation information of chain-relationship experts to amend the inconsistent scores and to adjust the differences in average and standard deviation between subgroups;(2)When evaluation activities tend to be on a large scale,it,s difficult to achieve evaluation one by one for experts and to realize interactive evaluation among all experts.In view of this,this paper proposes a two-stage sub-group decision making technology,which makes systematic sampling according to the consistency opinions in subgroups,and has interactive evaluation between all members.On this basis,to guarantee the reliability of the evaluation conclusion,it modifies the consistency of opinion of non-sample through test equating;(3)Considering time change and subjects vary,which is not tally with the facts,we put forward a dynamic group evaluation method.First of all,we make the differentiation decomposition for evaluation opinions,and then eliminate the effect of objects' level change to eonclusion due to time changes by measuring the conflict degree between subject of change and the subject of unchanged at the same period.Finally,we build a "bridge"through the information of unchanged subjects to implement all opinions comparable in different periods.Finally,the main results obtained in the dissertation are summarized,and several problems which need to be further perfected and studied are pointed in this thesis.