An Oversampling Method for Class Imbalance Problems on Large Datasets

Several oversampling methods have been proposed for solving the class imbalance problem. However, most of them require searching k-nearest neighbors to generate synthetic objects. This requirement makes time-consuming and therefore unsuitable large datasets. In this paper, an method problems that do not neighbors’ search is proposed. According our experiments on datasets with different sizes imbalance, at least twice as fast 8 fastest reported in literature while obtaining similar quality.