关键词: 综合评价, 群体评价, 客观修正, 数据仿真, 随机模拟

Abstract: Group evaluation is an important part of improving the scientific nature of decision-making and has a wide application to many fields at home and abroad. Whether the initial subjective evaluation data provided by group evaluators, such as the evaluation indicator value and the evaluation indicator weight value, is close to the true value is an essential factor in determining the quality of group evaluation. Therefore, from the perspective of some rational differences in the subjective data provided by group evaluators in group evaluation, this paper proposes an objective correction method to reduce the rational differences in group subjective evaluation data.
It is assumed that the evaluation data provided by the group evaluators has been rationally analyzed. The data does not contain evaluator preference information, and the existing differences are only caused by differences in the background, experience, knowledge, etc. of the group evaluators. The research method is as follows. Firstly, based on the idea of removing redundant data by factor analysis, this paper defines the connotation of the public factor of the true value of subjective evaluation data, and constructs an objective correction model. The application example shows that there are some differences and correlations between the method in this paper and the traditional one. Secondly, to illustrate the validity of the above model, the R language programming software is used to randomly generate virtual real values and test the model's validity. The stochastic simulation method is used to calculate the difference of superiority, and the different standard deviation of the superiority of the proposed method compared with the traditional one. Finally, considering that there are generally differences in the scenarios of actual group evaluation problems, in order to facilitate the evaluation developers to grasp the degree of the impact of evaluation parameters on the superiority degree at the beginning of the evaluation period, four categories of scenarios that can lead to changes in the superiority degree of the objective correction method are summarized based on the above steps. The four types of scenarios are: the scenario where the number of subjective evaluation data changes, the scenario where the number of group evaluators changes, the scenario where the true value interval changes, and the scenario where the rating error changes. Furthermore, the influence of scenario change in model parameters on the method's superiority and value superiority are analyzed systematically. It is found that the difference superiority and difference standard deviation superiority of this method are higher than the traditional one in each scenario.
The research results show that: (1)The method in this paper can remove redundant information contained in evaluation data to a certain extent, ameliorate the accuracy of evaluation data, reduce the cost of subjective method optimization data consumption, and improve evaluation efficiency. (2)As the amount of evaluation data increases or the difference in the true value of the initial estimate increases, the superiority data of this method increases. When the number of group evaluators increases or opinions become more dispersed, the superiority of this method will decrease. (3)The simulation test proves that this method has certain advantages over the traditional one. The superiority of this method is still high in different scenarios. This method has certain effectiveness and stability. In the future, we will conduct an in-depth research on improving the quality of group subjective evaluation data (parameter interaction, preference integration), thereby further improving the ability of this method for practical applications.

Key words: comprehensive evaluation, group evaluation, objective correction, data simulation, stochastic simulation