科技评价中专家评审粗差数据的处理及仿真

摘要

科技评价系统中,因评审专家主观认识上的偏差和评审项目客观存在的某种程度的争议性,导致评审数据不能客观地反应评审项目的真实情况,进而影响到评审工作的客观性与公正性.为解决上述问题,提出专家评价异常数据的修正模型及算法,首先利用小样本异常数据的稳健性处理方法进行评审异常数据检测;然后通过对异常数据进行分析,计算对应专家的评分偏差,并借助于熵值的概念计算项目的争议度.根据专家的偏差和项目的争议情况对评审异常数据进行修正,仿真结果表明了改进算法的有效性.%In science and technology evaluation,some scores assigned to the same object may deviate from the reality,due to the potential and subjective bias of reviewers and objective controversy of projects,thereby affecting the objectivity and imparity of the evaluation system.In this paper,the correction model and algorithm of abnormal data in evaluation system were presented.It used the robustness data processing method for small sample to detect the outlier firstly.As part of abnormal data analysis,it obtained the deviation of the reviewer and the controversy of the project by means of the concept of entropy.On the basis,the abnormal data were corrected.The model is shown to be effective by experiments on synthetic data sets.