Oversampling Method Based on Gaussian Distribution and K-Means Clustering

Learning from imbalanced data is one of the greatest challenging problems in binary classification, and this problem has gained more importance recent years. When class distribution imbalanced, classical machine learning algorithms tend to move strongly towards majority disregard minority. Therefore, accuracy may be high, but model cannot recognize instances minority classify them, leading many misclassifications. Different methods have been proposed literature handle imbalance problem, most are complicated simulate unnecessary noise. In paper, we propose a simple oversampling method based on Multivariate Gaussian K-means clustering, called GK-Means. The new aims avoid generating noise control imbalances between within classes. Various experiments carried out with six classifiers four methods. Experimental results different datasets show that GK-Means outperforms other improves classification performance as measured by F1-score Accuracy.