基于多准则排序融合的特征选择方法

摘要

针对模式分类中特征选择问题，为去除冗余特征，提高分类准确率，提出一种基于 ReliefF 算法、Fisher 比率算法和马氏距离算法的多准则排序融合的特征选择方法。动态结合上述3种单准则特征选择法的优点，实现对多个评价准则的综合利用。以 Ionosphere 标准数据集和高速列车转向架故障数据集为研究对象进行实验仿真，仿真结果表明，相比于单准则特征选择法，该方法能更有效地降低特征维数，具有更高的分类性能。%For the problem of feature selection in pattern classification,to remove redundant features and improve the classifica-tion accuracy,a multi-criterion feature ranking scheme was proposed.The proposed scheme was based on the fusion of a collec-tion of methods including ReliefF,Fisher ratio and Mahalanobis distance,the advantages of the three different evaluation criteria were combined,therefore the comprehensive utilization of the three criteria was realized.The four methods were simulated based on Ionosphere standard test and fault data of high-speed train,the experimental results show that,compared to the single criteria feature selection method,the proposed approach can effectively reduce feature dimension,and owns better classification perfor-mance.