Classification study for the imbalanced data based on Biased-SVM and the modified over-sampling algorithm
Liumei Zhang, Baoyu Tan, Tianshi Liu, Xiaoqun Sun · Journal of Physics Conference Series · 2019
By combining the modified Random-SMOTE oversampling algorithm with the Biased-SVM classification method, this paper has proposed an improved classification approach for the imbalanced data sets. This algorithm is able to cluster the minority samples, and ensures the support vectors as the parent samples according to the distances between the cluster centers in the minority class and the majority class center. It then could generate the new samples for the minority class. The experiment is conducted upon the five imbalanced data sets from UCI data set and the proposed algorithm is compared with other algorithms. The experimental results show that the improved algorithm has significant classification effect for the imbalanced data sets.