期刊评价中TOPSIS的漏洞研究——权重单调性

Study on TOPSIS's Vulnerability in Journal Evaluation-weights Monotony

首先提出非线性评价法权重单调性问题,即在多属性评价中,在其他指标权重不变的情况下,其中一个指标权重增加,那么评价结果也必须增加,并设计出采用传统回归和岭回归计算模拟权重进而对权重的单调性进行检验的方法。基于JCR2012物理学期刊数据,以总被引频次和影响因子为例对TOPSIS评价方法的研究表明:TOPSIS评价法权重并不具有正向单调性,只能用于等权重评价,加权TOPSIS法存在漏洞。研究方法也可广泛用于非线性评价方法的权重单调性检验。

This paper firstly proposed the problem of weights monotony of non-linear evaluation methods, namely in a multi-attribute e-valuation, with other indexes＇ weights remaining unchanged and only one index＇s weight being increased, the evaluation results must also increase. Then,the paper designed the methods using traditional regression and ridge regression to calculate the simulant weights and the monotony of the weights were thentested. Based on the data of JCR2012 physics journals, the paper took the total citations and impacted factors as an example to study the evaluation method of TOPSIS and showed that： The evaluation method of TOPSIS does not have the positive monotony, and only can be used for equal weighted evaluation;the method of weighted TOPSIS has loopholes. The method used in this paper can also be widely used in the weighted monotonic test of non-linear evaluation methods.