Research Article
A Novel Ensemble Method for Imbalanced Data Learning: Bagging of Extrapolation-SMOTE SVM
(1) Input the whole dataset, the number of SVMs-, the oversampling ratio for | Extrapolation Borderline-SMOTE . | (2) Train on the original data set to fit soft margin SVM by choosing a proper | kernel and hyper-parameter in cross-validation and identify the support vectors | belonging to the minority | (3) For from 1 to : | (4) Bootstrap on the to obtain the sampling result and | not sampled in the turn | (5) Get the union set as and operate extrapolation | borderline-SMOTE with the sampling ratio on it. | (6) and synthetic samples are united as training data set to obtain | soft margin SVM with hyper-parameter chosen by validating the | performance on . | (7) Output the ensemble of SVMs |
|