Abstract
To deal with classification problems, support vector machines (SVMs) are utilized in a wide variety of applications as effective and powerful supervised learning paradigm. However, the efficacy and outcomes of an SVM-based classification model is influenced by the proper selection of SVM parameters in addition to the nature of the datasets. Therefore, the purpose of this work is to enrich the efficacy of the SVMs based on simultaneous optimization of the parameters and feature weighting of these models. In this paper, an improved evolutionary variant of competitive swarm optimizer (CSO) is proposed to evolve the parameters of SVMs and optimize the weights of features. Simulations and experiments are performed based on various datasets from UCI repository to investigate the effectiveness of the proposed hybrid CSO-based SVM model versus genetic algorithm, particle swarm optimizer and the classical grid-based search. Results and analysis reveal that the proposed crossover-based mechanism inside CSO has improved the classification capabilities of the hybrid CSO-SVM technique.