基于相关性分析的特征选择方法研究

A study on methods of feature selection based on the correlation analysis

讨论了特征相关性的大小与分类能力之间的联系,提出了一种以相关系数和为判据的特征自动选择方法,并将它应用到地下核爆炸与天然地震的自动识别中,实验结果表明该方法所提取的特征子集具有很好的分类能力。

The connection between the correlation and the classify ability of feature is studied. A method of feature automatic selection based on the criterion of correlation coefficient is proposed. And the feature subclass selected by using this method is proved to have powerful classification ability by the good results of experiments to automatic discriminate the underground nuclear explosion and natural earthquake.