Moeinoddin Sheikhottayefe - Amirkabir University of TechnologyHafez Ave, Tehran, Iran

Support Vector Machine (SVM) is a wellknown machine learning method commonly used to address classification and regression tasks. The bases of SVM are structural risk minimization and the statistical learning theory. SVM’s efficacy and classification accuracy are associated with setting its parameters and the subset of selected features. This paper provides a method for choosing optimum features and optimizing SVM parameters at the same time, based on an improvement over a novel metaheuristic algorithm called Slime Mold Algorithm (SMA). This improvement is based on incorporating Levy Flight into the original SMA. Besides implementing as a tuner of SVM’s parameters, the Advanced SMA (ASMA) selects the most appropriate subset of attributes for the functioning of the Support Vector Machine. As the experimental data indicates, reduction of the number of attributes while ensuring excellent prediction accuracy is possible by utilizing our proposed method. Index Terms—Advanced Slime Mold Optimizer, SMA , SVM, Levy Flight ,

Index Terms—Advanced Slime Mold Optimizer, SMA , SVM, Levy Flight , Support Vector Machines, Optimization,