Formulating new enhanced pattern classification algorithms based on ACO-SVM

• Main Authors: Alwan, Hiba Basim, Ku-Mahamud, Ku Ruhana
• Format: Article
• Language: English
• Published: North Atlantic University Union NAUN 2013
• Subjects: QA76 Computer software
• Online Access: http://repo.uum.edu.my/9845/1/p.pdf; http://repo.uum.edu.my/9845/; http://www.scimagojr.com/journalsearch.php?q=18100156703&tip=sid&clean=0

This paper presents two algorithms that integrate new Ant Colony Optimization (ACO) variants which are Incremental Continuous Ant Colony Optimization (IACOR) and Incremental Mixed Variable Ant Colony Optimization (IACOMV) with Support Vector Machine (SVM) to enhance the performance of SVM.The first algorithm aims to solve SVM model selection problem. ACO originally deals with discrete optimization problem.In applying ACO for solving SVM model selection problem which are continuous variables, there is a need to discretize the continuously value into discrete values.This discretization process would result in loss of some information and hence affects the classification accuracy and seeking time.In this algorithm we propose to solve SVM model selection problem using IACOR without the need to discretize continuous value for SVM.The second algorithm aims to simultaneously solve SVM model selection problem and selects a small number of features.SVM model selection and selection of suitable and small number of feature subsets must occur simultaneously because error produced from the feature subset selection phase will affect the values of SVM model selection and result in low classification accuracy.In this second algorithm we propose the use of IACOMV to simultaneously solve SVM model selection problem and features subset selection.Ten benchmark datasets were used to evaluate the proposed algorithms.Results showed that the proposed algorithms can enhance the classification accuracy with small size of features subset.