A robust multiobjective Harris’ Hawks Optimization algorithm for the binary classification problem[Formula presented]

Dokeroglu T., Deniz A., Kiziloz H. E.

Knowledge-Based Systems, cilt.227, 2021 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 227
  • Basım Tarihi: 2021
  • Doi Numarası: 10.1016/j.knosys.2021.107219
  • Dergi Adı: Knowledge-Based Systems
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Applied Science & Technology Source, Computer & Applied Sciences, INSPEC, Library and Information Science Abstracts, Library, Information Science & Technology Abstracts (LISTA)
  • Anahtar Kelimeler: Binary classification, Multiobjective optimization, Feature selection, Harris' Hawks optimization, ARTIFICIAL BEE COLONY, FEATURE-SELECTION, DIFFERENTIAL EVOLUTION, MACHINE, SEARCH
  • Orta Doğu Teknik Üniversitesi Adresli: Evet

Özet

© 2021 Elsevier B.V.The Harris’ Hawks Optimization (HHO) is a recent metaheuristic inspired by the cooperative behavior of the hawks. These avians apply many intelligent techniques like surprise pounce (seven kills) while they are catching their prey according to the escaping patterns of the target. The HHO simulates these hunting patterns of the hawks to obtain the best/optimal solutions to the problems. In this study, we propose a new multiobjective HHO algorithm for the solution of the well-known binary classification problem. In this multiobjective problem, we reduce the number of selected features and try to keep the accuracy prediction as maximum as possible at the same time. We propose new discrete exploration (perching) and exploitation (besiege) operators for the hunting patterns of the hawks. We calculate the prediction accuracy of the selected features with four machine learning techniques, namely, Logistic Regression, Support Vector Machines, Extreme Learning Machines, and Decision Trees. To verify the performance of the proposed algorithm, we conduct comprehensive experiments on many benchmark datasets retrieved from the University of California, Irvine (UCI) Machine Learning Repository. Moreover, we apply it to a recent real-world dataset, i.e., a Coronavirus disease (COVID-19) dataset. Significant improvements are observed during the comparisons with state-of-the-art metaheuristic algorithms.