Building a hybridised meta-heuristic optimisation algorithm for efficient cluster analysis

作     者：Kumar, D. Pradeep Sowmya, B.J. Kanavalli, Anita Cornelio, Varun Dsouza, Jaison Pravith Memon, Wasim Prashanth, P. 

作者机构：Department of Computer Science and Engineering M.S. Ramaiah Institute of Technology Bangalore India 

出 版 物：《International Journal of Business Intelligence and Data Mining》 (Int. J. Bus. Intell. Data Min.)

年 卷 期：2022年第22卷第1-2期

页      面：170-222页

核心收录：

学科分类：12[管理学] 02[经济学] 0710[理学-生物学] 1205[管理学-图书情报与档案管理] 120202[管理学-企业管理（含：财务管理、市场营销、人力资源管理）] 0202[经济学-应用经济学] 1004[医学-公共卫生与预防医学(可授医学、理学学位)] 1202[管理学-工商管理] 1002[医学-临床医学] 020205[经济学-产业经济学] 1001[医学-基础医学(可授医学、理学学位)] 0835[工学-软件工程] 0714[理学-统计学（可授理学、经济学学位）] 0836[工学-生物工程] 0701[理学-数学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

主　　题：Cluster analysis 

摘      要：Nature-inspired algorithms are a relatively recent field of meta-heuristics introduced to optimise the process of clustering unlabelled data. In recent years, hybridisation of these algorithms has been pursued to combine the best of multiple algorithms for more efficient clustering and overcoming their drawbacks. In this paper, we discuss a novel hybridisation concept where we combine the exploration and exploitation processes of the vanilla bat and vanilla whale algorithm to develop a hybrid meta-heuristic algorithm. We test this algorithm against the existing vanilla meta-heuristic algorithms, including the vanilla bat and whale algorithm. These tests are performed on several single objective CEC functions to compare convergence speed to the minima coordinates. Additional tests are performed on several real-life and artificial clustering datasets to compare convergence speeds and clustering quality. Finally, we test the hybrid on real-world cases with unlabelled clustering data, namely a credit card fraud detection dataset, and a COVID-19 diagnosis dataset, and end with a discussion on the significance of the work, its limitations and future scope. © 2023 Inderscience Enterprises Ltd.