ReliefF-MFO多标签特征选择算法  

ReliefF-MFO multi-label feature selection algorithm

作　　者：何牧宇 周晖[1] HE Mu-yu;ZHOU Hui(School of Electronics and Information,Nantong University,Nantong 226019,China)

机构地区：[1]南通大学电子信息学院

出　　处：《计算机工程与设计》2019年第12期3469-3473,共5页Computer Engineering and Design

基　　金：国家自然科学基金项目(61501264)

摘　　要：为解决启发式算法在多标签特征选择中可以达到较好效果但效率很低的问题,提出一种基于启发式算法的混合特征选择算法。使用ReliefF方法去除不相关特征,采用MFO算法进行特征子集寻优,提高分类器性能的同时达到较高的效率。将所提方法应用于多个典型多标签数据集分类问题并与现有启发式特征选择方法进行对比,实验结果表明了所提算法的有效性。Aiming at the problem that heuristic algorithm achieves acceptable effects with low efficiency in multi-label feature selection,a hybrid feature selection algorithm based on heuristic algorithm was proposed.ReliefF method was used to remove irrelevant features,and MFO algorithm was used to optimize feature subset,to improve the performance of classifier and achieve high efficiency at the same time.The proposed method was applied to the classification of multiple typical multi-label datasets and compared with the existing heuristic feature selection method.The experimental results show that the proposed algorithm is effective.

关 键 词：多标签分类 多标签特征选择 混合式特征选择 飞蛾火焰优化算法 RELIEFF算法 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]