2026-06-27T00:03:33Z https://meral.edu.mm/oai

oai:meral.edu.mm:recid/4794 2021-12-13T03:35:08Z 1582963302567:1597824273898 user-ucsy

Modified-MCA Based Feature Selection Model for Classification Khaing, Myo Kham, Nang Saing Moon A central problem in machine learning isidentifying a representative set of features fromwhich to construct a classification model for aparticular task. A good feature set that containshighly correlated features with the class not onlyimproves the efficiency of the classificationalgorithms but also improve the classificationaccuracy. Modified-Multiple CorrespondenceAnalysis (M-MCA or MCA with GeometricalRepresentation) explores the correlation betweendifferent features and classes to score thefeatures for feature selection. The dependencebetween a feature and a class is measured by aderived value from χ2 distance called the p-value.It is a standard measure of the reliability of arelation and is examined by p-value. The smallerthe p-value, the higher the possibility of thecorrelation between a feature and a class is true.In this paper, the conventional confidenceinterval of Multiple Correspondence Analysis(MCA) is modified to get smaller p-value and bemore reliable. To evaluate the performance ofproposed Modified-MCA, experiments arecarried out on benchmark datasets identified andprovided by WEKA and UCI repository. In theexperiments, Naïve Bayes, Decision Table andJRip are used as the classifiers. The proposedModified-MCA demonstrates promising resultsand performs better than well-known featureselection, MCA. The results show that theproposed method outperforms in terms ofclassification accuracy and reduces the size offeature subspace significantly. 2011-05-05 http://hdl.handle.net/20.500.12678/0000004794 https://meral.edu.mm/records/4794