基于乌鸦搜索算法的新型特征选择算法

A Novel Feature Selection Algorithm Based on Crow Search Algorithm

基于元启发式算法乌鸦搜索算法(CrSA),提出一种改进的基于乌鸦搜索算法的特征选择算法(IFSCrSA),以解决目前特征选择问题中存在的不足.通过与传统的机器学习特征选择算法和基于进化计算的特征选择算法进行比较,结果表明,IFSCrSA能在数据集中选择辨识度较强的特征,不仅大幅度降低了特征子集的规模,而且提高了分类准确率。

Based on the meta-heuristic algorithm: crow searc h algorithm (CrSA),we proposed an improved feature selection based on crow search algorithm (IFSCrSA) to solve the shortcomings of the current feature selection problem. Compared with traditional machine learning feature selection algorithms and feature selection algorithm based on evol utionary computing. The results show that IFSCrSA can select features with stron g recognition in the data set,which not only greatly reduces the size of the feature subset,but also improves the classification accuracy.