序比例诱导分段无量纲化方法及其影响因素

为在评价数据的无量纲化处理过程中兼顾异常值情形，并减弱因异常值的存在而导致的无量纲化后数据之间分布不均衡的问题，在极值处理法的基础上，提出了一种非线性的无量纲化方法，即序比例诱导分段无量纲化方法。该方法以原始指标值的排序百分比为诱导变量对原始指标值进行分段，并在各区段内以极值处理法为基础分别进行无量纲化处理。通过性质分析，发现该方法能够较大程度地提升无量纲化结果对“总量恒定性”性质的满足程度。此外，采用模拟仿真的方法发现，该无量纲化方法对异常值具有较好的抗干扰性、且随着分段层级的增加其对异常值的敏感程度越来越低等主要结论。

A Segmented Dimensionless Method Induced by Ranking Percentage and Its Influencing Factors

In order to consider the situation of outliers in dimensionless processing of the evaluation data, and to reduce the imbalance distribution between the dimensionless data caused by outliers, this paper proposes a novel nonlinear dimensionless method, namely, the  segmented dimensionless method induced by ranking percentage, based on the extremum method, which takes the ranking percentage of initial data as the inducing variable to segment the initial data, and performs dimensionless processing in different intervals based on the extremum method. The property analysis shows that this method can greatly improve the satisfaction degree of the dimensionless data to the property of the “constant total”. Moreover, this dimensionless method has a better anti-interference ability for outliers, and as the segment increases, its sensitivity to outliers decreases.