首页|Boosted Spider Wasp Optimizer for High-dimensional Feature Selection

Boosted Spider Wasp Optimizer for High-dimensional Feature Selection

扫码查看
原文链接
  • 万方数据
With the increasing dimensionality of the data,High-dimensional Feature Selection(HFS)becomes an increasingly dif-ficult task.It is not simple to find the best subset of features due to the breadth of the search space and the intricacy of the interactions between features.Many of the Feature Selection(FS)approaches now in use for these problems perform sig-nificantly less well when faced with such intricate situations involving high-dimensional search spaces.It is demonstrated that meta-heuristic algorithms can provide sub-optimal results in an acceptable amount of time.This paper presents a new binary Boosted version of the Spider Wasp Optimizer(BSWO)called Binary Boosted SWO(BBSWO),which combines a number of successful and promising strategies,in order to deal with HFS.The shortcomings of the original BSWO,including early convergence,settling into local optimums,limited exploration and exploitation,and lack of population diversity,were addressed by the proposal of this new variant of SWO.The concept of chaos optimization is introduced in BSWO,where initialization is consistently produced by utilizing the properties of sine chaos mapping.A new convergence parameter was then incorporated into BSWO to achieve a promising balance between exploration and exploitation.Multiple exploration mechanisms were then applied in conjunction with several exploitation strategies to effectively enrich the search process of BSWO within the search space.Finally,quantum-based optimization was added to enhance the diversity of the search agents in BSWO.The proposed BBSWO not only offers the most suitable subset of features located,but it also lessens the data's redundancy structure.BBSWO was evaluated using the k-Nearest Neighbor(k-NN)classifier on 23 HFS problems from the biomedical domain taken from the UCI repository.The results were compared with those of traditional BSWO and other well-known meta-heuristics-based FS.The findings indicate that,in comparison to other competing techniques,the proposed BBSWO can,on average,identify the least significant subsets of features with efficient classification accuracy of the k-NN classifier.

High-dimensional featuresSWO algorithmFeature selectionOptimizationMachine learning

Elfadil A.Mohamed、Malik Sh.Braik、Mohammed Azmi Al-Betar、Mohammed A.Awadallah

展开 >

Artificial Intelligence Research Center(AIRC),College of Engineering and Information Technology,Ajman University,346,Ajman,United Arab Emirates

Department of Computer Science,Al-Balqa Applied University,Al-Salt 19117,Jordan

Department of Information Technology,Al-Balqa Applied University,Irbid 21510,Jordan

Department of Computer Science,Al-Aqsa University,Gaza 4051,Palestine

Artificial Intelligence Research Center(AIRC),Ajman University,346,Ajman,United Arab Emirates

展开 >

2024

10.1007/s42235-024-00558-8

仿生工程学报(英文版)
吉林大学

仿生工程学报(英文版)

CSTPCDEI
影响因子：0.837
ISSN：1672-6529
年,卷(期)：2024.21(5)