Imbalanced Classification Based on Over-sampling and Feature Selection
Qian Wang · 2020
The imbalanced classification is an important problem in the field of machine learning. In the imbalanced classification problem, correct classification for the minority classes is usually hard. This paper presents an improved approach, named AdaBoost-SVM-OBFS. Firstly, the ReliefF algorithm is used to select more classification-related features; Secondly, we augment the minority samples using Boosting algorithm and the over-sampling technique; Finally the new generated samples are added to the original training set to retrain the SVM. Under the evaluation indexes of AUC, F-Measure and FP Rate, the AdaBoost-SVM-OBFS approach is compared with the AdaBoost-SVM on six imbalanced data sets. The experimental results demonstrate that the proposed approach is valid for the classification of imbalanced data.