Optimal SVM Parameters Estimation Using Chaotic Accelerated Particle Swarm Optimization for Genetic Data Classification

Microarray studies and gene expression analysis havereceived tremendous attention over the last few years and providemany promising paths toward the understanding of fundamentalquestions in biology and medicine. High throughput biological dataneed to be processed, analyzed, and interpreted to address problemsin life sciences. Feature selection (FS) and clustering are amongthe methods used in the statistical analysis of these data types.Support Vector Machine (SVM) is very popular and powerfulamong learning methods. In general, SVM has a wide range ofapplications in pattern recognition and nonlinear fitting. Kernelfunctions have an important role in classification capability ofSVM. In this study, at first, selection of appropriate and efficientfeatures is performed using Fisher ranking method consideringclassification power. After that, balancing is performed betweendifferent classes based on the removal of samples of the majorityclass, which are far from the decision boundary. Optimization ofSVM classifier parameters including penalty factor C and RadialBasis Function (RBF) parameter

Keywords:

Microarray , support vector machine , feature selection , Fisher ranking method , chaotic accelerated particle swarmoptimization