摘要
马田系统是一种产生于质量工程学领域中的模式识别方法,该方法能够在不平衡数据环境中进行异常值识别和特征选择,并且不需要任何数据分布假设,具有样本需求量小、原理简单和易于操作等优点.为更好地促进马田系统理论及应用研究,首先对马田系统的异常值识别和特征选择原理进行介绍;然后从马氏距离、信噪比、马氏空间、特征选择、阈值、数据环境和应用等7个方面,梳理马田系统的研究进展;最后,对马田系统研究进展进行总结,指出马田系统仍存在的问题以及未来可能的研究方向.
As a pattern recognition method generated in the quality engineering field, the Mahalanobis-Taguchi system(MTS) can carry out outlier recognition and feature selection in an imbalanced data environment without any assumption of data distribution. It has the advantages such as small samples, simple principle and easy operation. In order to better promote the theoretical and applied research of the MTS, this paper first introduces the basic principles of outlier recognition and feature selection of the MTS, and then reviews the research progress in terms of Mahalanobis distance,signal to noise ratio, Mahalanobis space, feature selection, threshold, data environment, and the application of the MTS.Finally, this paper summarizes the research progress and proposes a detailed analysis of the future possible research directions of the MTS.
作者
常志朋
CHANG Zhi-pengy(School of Business,Anhui University of Technology,Maanshan 243002,China;Institute of Anhui’s Innovation Driving and Development,Anhui University of Technology,Maanshan 243002,China)
出处
《控制与决策》
EI
CSCD
北大核心
2019年第12期2505-2516,共12页
Control and Decision
基金
国家自然科学基金面上项目(71673001)
安徽省高校优秀青年人才支持计划重点项目(gxyqZD2017040)
关键词
马田系统
马氏距离
马氏空间
信噪比
正交表
特征选择
Mahalanobis-Taguchi system
Mahalanobis distance
Mahalanobis space
signal to noise ratio
orthogonal array
feature selection
