一种基于组策略的过滤式特征选择算法

MRMR算法具有快速、高效等优势，在处理高维数据方面较为流行。基于此，提出一种基于组策略的MRMR改进算法（MRMRE），该算法不仅考虑单个特征属性的相关性与冗余性，同时针对特征组间的相互关系进行研究。算法以MRMR算法为框架，以CCA作为度量基准，选择SVMs作为基分类器，使其特征选择效果提升。在UCI机器学习数据库中图像与基因序列数据集上的大量实验表明：与MRMR算法相比，所提出的算法其特征选择结果具有更高的结果稳定性与分类精度。

A filter feature selection algorithm based on the group policy

The MRMR algorithm is fast and effective, and it is popular in the handling of high-dimension data. Motivated by this, this paper proposed a refined MRMR algorithm (MRMRE) based on the group mechanism. To improve the results of feature selection, this algorithm not only considers the relationship between features, but also considers the relationship between feature groups. This algorithm takes the MRMR algorithm as the frame, uses CCA as the measure and selects SVMs as the base classifier. Massive experiments conducted on generous images and gene sequence data sets in the machine learning database from UCI show that the proposed algorithm has higher result stability and classification precision in feature selection compared to the MRMR algorithm.