Neural Network-Based Undersampling Techniques

Machine learning models have gained popularity nowadays for their potential to solve real-life issues when trained on pertinent data. In many cases, the data are class imbalanced and hence corresponding machine tend perform poorly metrics like precision, recall, AUC, F1, G-mean score. Since imbalance issue poses serious challenges performance of models, a multitude research works addressed this issue. Two common data-based sampling techniques mostly been proposed-undersampling majority oversampling minority class. article, we focus former approach. We propose two novel algorithms that employ neural network-based approaches remove samples found reside in vicinity samples, thereby undersampling (or alleviate) delineate proposed then test some publicly available datasets. compare our other popular algorithms. Finally, conclude outperform most existing metrics.