A Cluster Based Classification of Imbalanced Data with Overlapping Regions Between Classes
Conference proceedings article
Authors/Editors
Strategic Research Themes
No matching items found.
Publication Details
Author list: Chujai P., Choomboon K., Chaiyakhan K., Kerdprasop K., Kerdprasop N.
Publication year: 2017
Volume number: 2227
Start page: 353
End page: 358
Number of pages: 6
ISBN: 9789881404732
ISSN: 2078-0958
Languages: English-Great Britain (EN-GB)
Abstract
Classifying imbalanced data is a significant challenge for machine learning algorithms. Difficulty is due to the fact that data in the minority class can easily be overshadowed by the much larger number of instances in the majority class. The overall classification accuracy may be high, but the recognition of data instances in the minority class are normally unacceptable when applying standard algorithms. Therefore, this research proposes a technique for handling the imbalanced classification problem. We solve the imbalanced classification problem by performing separation of the imbalanced data into overlapped and non-overlapped regions between majority and minority classes. After the separation, data were clustered based on Euclidean distance consideration. Each cluster, then, has its own classification model. To predict the future event, closest distance scheme from all models has been applied. The experimental results show that the proposed technique modeling with the SVM using linear kernel function yields the best performance in classifying minority data.
Keywords
Imbalanced data classification, Overlapping region, SVM with linear kernel
Contact Information