期刊
KNOWLEDGE-BASED SYSTEMS
卷 260, 期 -, 页码 -出版社
ELSEVIER
DOI: 10.1016/j.knosys.2022.110111
关键词
Feature selection; Artificial immune algorithm; Hybrid search; High-dimensional data; Classification
This paper proposes a hybrid feature selection method based on artificial immune algorithm optimization (HFSIA) to solve the feature reduction problem of high-dimensional data. Experimental comparisons show that HFSIA has comparable computational cost to classical feature selection methods known for their speed, while achieving higher average classification accuracy and feature reduction rate on benchmark datasets.
For high-dimensional data, the traditional feature selection method is slightly inadequate. At present, most of the existing hybrid search methods have problems of high computational cost and unsatisfac-tory feature reduction rate. In this paper, a hybrid feature selection method based on artificial immune algorithm optimization (HFSIA) is proposed to solve the feature reduction problem of high-dimensional data. This method combines the filter method with the metaheuristic-based search strategy more effectively. Inspired by biological research results, the method introduces a lethal mutation mechanism and a Cauchy mutation operator with adaptive adjustment factors to improve the search performance of the algorithm. In addition, this method introduces an adaptive adjustment factor in the population update stage to improve the problem of insufficient diversity of the original algorithm. The effective combination of these mechanisms enables the algorithm to obtain better search capability at a lower computational cost. Experimental comparisons with 23 state-of-the-art feature selection methods are conducted on 22 high-dimensional benchmark datasets. The results show that the computational cost of HFSIA is comparable to 5 classical feature selection methods known for their speed. Moreover, it achieves a higher average classification accuracy than 18 hybrid feature selection methods reported in the latest literature with the best feature reduction rate.(c) 2022 Elsevier B.V. All rights reserved.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据