整体与个体视角下非线性学术评价伪权重研究——以加权TOPSIS法为例

Research on Pseudo Weight of Nonlinear Academic Evaluation from the Overall Perspective and Individual Perspective:Taking Weighted TOPSIS Method as an Example

在非线性学术评价中,存在设计权重与实际权重的偏离即伪权重问题,使得以专家权重为主的设计权重在评价中失效,从整体与个体视角进行系统分析十分必要。文章在理论分析的基础上,以加权TOPSIS和JCR2019经济学期刊为例,基于岭回归与偏最小二乘法拟合实际权重,同时采用极大值标准化与sigmoid函数标准化,比较其对实际权重的影响,并进行伪权重的相关问题分析。研究结果表明:指标标准化方法与权重拟合方法是影响整体伪权重的重要因素;采用sigmoid函数标准化可以降低整体伪权重问题;非线性评价整体伪权重问题在理论上无法消除;基于岭回归与偏最小二乘法估计实际权重具有较好的精度;个体伪权重问题在非线性评价中同样存在;采用sigmoid函数可以有效降低个体伪权重问题;个体指标发展不均衡产生的伪权重问题一定程度上值得鼓励。

In the nonlinear academic evaluation,there is a deviation between the design weight and the actual weight,that is,the pseudo weight problem,which makes the design weight based on expert weight invalid in the evaluation.It is necessary to systematically analyze from the overall and individual perspectives.On the basis of theoretical analysis,taking weighted TOPSIS and JCR2019 economics journals as examples,this paper takes the ridge regression and partial least squares method to fit the actual weights.At the same time,the maximum standardization and sigmoid function standardization are used to compare their effects on the actual weights,and the related problems of pseudo weights are analyzed.The results show that the index standardization method and weight fitting method are important factors affecting the overall pseudo weight;using sigmoid function standardization can reduce the overall pseudo weight problem;the overall pseudo weight of nonlinear evaluation cannot be eliminated theoretically;based on ridge regression and partial least squares method to estimate the actual weight has good accuracy;individual pseudo weight problems exist in nonlinear evaluation;using sigmoid function can effectively reduce the individual pseudo weight problem;the pseudo weight problem caused by the unbalanced development of individual indicators is worth encouraging to some extent.