2024-03-28T15:29:12Z
https://meral.edu.mm/oai
oai:meral.edu.mm:recid/4447
2022-03-24T23:12:14Z
1582963302567:1597824273898
user-ucsy
Effective Feature Selection for Preprocessing Step of Classification Using Modified-MCA
Khaing, Myo
Kham, Nang Saing Moon
A novel metric that integrates the correlationand reliability information between each featureand each class obtained from MultipleCorrespondence Analysis (MCA) is currently thepopular solution to score the features for featureselection. However, it has the disadvantage thatp-value which examines the reliability isconventional confidence interval. The main goalof this paper is to introduce a new classifierindependent (filter-based) feature selectionmethod, Modified Multiple CorrespondenceAnalysis (Modified-MCA) which is designed tomodify MCA, improving the reliability. Theefficiency and effectiveness of proposed methodis demonstrated through extensive comparisonswith MCA and other feature selection methods,using five benchmark datasets provided byWEKA and UCI repository. Naïve Bayes,Decision Tree and JRip are used as theclassifiers. The classification results, in terms ofclassification accuracy and size of featuresubspace, show that the proposed ModifiedMCA outperforms three other feature selectionmethods, MCA, Information Gain, and Relief
2012-02-28
http://hdl.handle.net/20.500.12678/0000004447
https://meral.edu.mm/records/4447