A Modified Group Search Optimizer for Feature Selection and Parameter Determination of SVM
K. Joshil Raj1, S Siva Sathya2, Kalyan Nandi3
1K. Joshil Raj, Department of Computer Science,Pondicherry University, Puducherry, India.
2Dr. S Siva Sathya, Department of Computer Science, Pondicherry University, Puducherry, India.
3Kalyan Nandi, Department of Computer Science, Pondicherry University, Puducherry, India.
Manuscript received on April 14, 2015. | Revised Manuscript received on April 26, 2015. | Manuscript published on March 05, 2015. | PP: 32-36 | Volume-5, Issue-2, May 2015. | Retrieval Number: B2590055215/2015©BEIESP
Open Access | Ethics and Policies | Cite
©The Authors. Published By: Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/
Abstract: Support vector machine (SVM) is a popular pattern classification method with many diverse applications. Group Search Optimizer (GSO) is a new population based optimization algorithm inspired by animal searching behavior for developing optimum searching strategies to find out solutions for continuous optimization problems. This paper presents an experimental analysis of modifications to classical GSO & studies its effects on a GSO-SVM hybrid combination for feature selection and kernel parameters optimization. In the proposed algorithm, three modifications are introduced over classical GSO to improve its global search mechanism. The quality and effectiveness of the proposed methodology has been evaluated on standard machine learning datasets.
Keywords: Evolutionary algorithm; Group Search Optimizer; GSO; Support Vector Machine; Machine learning; Feature Selection; Kernel parameters.