Genetic Algorithms Designed for Solving Support Vector Classifier

The support vector machine (SVM) is a newly developed approach in data mining. Using SVM, the classification and the regression problems can be converted into optimization problems with linear constraints. In this paper, a genetic algorithm employing mixed coding scheme, dual evolutionary iteration and some specially designed operators is proposed to perform comprehensive optimization for SVM and to process the constraints at the same time. The experiments upon several benchmark datasets prompts that the proposed algorithm performs better comparing to some other classification methods.