HWMWOA: A Hybrid WMA-WOA Algorithm with Adaptive Cauchy Mutation for Global Optimization and Data Classification

作     者：Zhang, Jiali Li, Haichan Parizi, Morteza Karimzadeh 

作者机构：Guangzhou Coll Technol & Business Coll Engn Guangzhou 510800 Guangdong Peoples R China Software Engn Inst Guangzhou Dept Network Technol Guangzhou 510990 Guangdong Peoples R China Tech & Vocat Univ TVU Fac Shahid Chamran Dept Comp Engn Kerman Branch Kerman Iran 

出 版 物：《INTERNATIONAL JOURNAL OF INFORMATION TECHNOLOGY & DECISION MAKING》 (国际信息技术与决策杂志)

年 卷 期：2023年第22卷第4期

页      面：1195-1252页

核心收录：

学科分类：12[管理学] 1201[管理学-管理科学与工程(可授管理学、工学学位)] 08[工学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：2020 Guangzhou College of Technology and Business School-level Quality Engineering Construction Project "Big Data Course Teaching Reform Based on Chaoxing Fanya Network Teaching Platform-"Data Analysis and Mining Practice (Python)' as an Example" [ZL20201243] 

主　　题：Whale Optimization Algorithm Woodpecker Mating Algorithm Cauchy mutation Support Vector Machine data classification 

摘      要：Combinatorial metaheuristic optimization algorithms have newly become a remarkable domain for handling real-world and engineering design optimization problems. In this paper, the Whale Optimization Algorithm (WOA) and the Woodpecker Mating Algorithm (WMA) are combined as HWMWOA. WOA is an effective algorithm with the advantage of global searching ability, where the control parameters are very less. But WOA is more probable to get trapped in the local optimum points and miss diversity of population, therefore suffering from premature convergence. The fundamental goal of the HWMWOA algorithm is to overcome the drawbacks of WOA. This betterment includes three basic mechanisms. First, a modified position update equation of WMA by efficient exploration ability is embedded into HWMWOA. Second, a new self-regulation Cauchy mutation operator is allocated to the proposed hybrid method. Finally, an arithmetic spiral movement with a novel search guide pattern is used in the suggested HWMWOA algorithm. The efficiency of the suggested algorithm is appraised over 48 test functions, and the optimal outcomes are compared with 15 most popular and newest metaheuristic optimization algorithms. Moreover, the HWMWOA algorithm is applied for simultaneously optimizing the parameters of SVM (Support Vector Machine) and feature weighting to handle the data classification problem on several real-world datasets from the UCI database. The outcomes prove the superiority of the suggested hybrid algorithm compared to both WOA and WMA. In addition, the results represent that the HWMWOA algorithm outperforms other efficient techniques impressively.