CSO-Based Feature Selection and Parameter Optimization for Support
Kuan-Cheng Lin · 2009
This research constructs the eSO+SVMmodel for data classification through integrating cat swam optimization intoSVMclassifier. There are two factors (i.e. feature selection and parameter determination)of classification problems will mainly discuss in this study. The objectivesoffeature selection are to reduce number offeatures and remove irrelevant, noisy and redundant data. Besides, the parameter optimization for training can improve classification performance. Hence, the optimal feature subset and kernel parameter are applied toSVM classifier for reducing the computational time in an acceptable classification accuracy. Furthermore, the classification accuracy is increased. The different classes and types in Uel machine learning repository is used to evaluate the classification accuracyofthe proposedeSO+SVMand GA+SVM methods.. Experimental results show the effectiveness of the proposed eSO+SVM method for solving data classification problems.