Datasets in real world are often predominately composed of normal examples with only a small percentage of interesting or abnormal examples. A new approach is applied in this paper to address the imbalance problem by combining SVM ensembles and resampling method. Through empirical analysis, researchers cluster majority classes by k-means algorithm into subclass which decreases the imbalance ratio. Additionally, they use resampling method which concludes oversampling and undersampling techniques to deal with the problem of long training time and low training efficiency in SVM ensembles. Experimental results show that the SVM ensembles with resampling method outperform individual SVM classifiers. The proposed combination approach can effectively solve the imbalance problem. © 2013 Springer Science+Business Media New York.
CITATION STYLE
Chen, X., Zhang, Y., & Wu, K. (2013). Combination approach of svm ensembles and resampling method for imbalanced datasets. In Lecture Notes in Electrical Engineering (Vol. 236 LNEE, pp. 641–649). Springer Verlag. https://doi.org/10.1007/978-1-4614-7010-6_72
Mendeley helps you to discover research relevant for your work.