Solving the class imbalance problem using a counterfactual method for data augmentation

Learning from class imbalanced datasets poses challenges for many machine learning algorithms. Many real-world domains are, by definition, virtue of having a majority that naturally has more instances than its minority (e.g., genuine bank transactions occur much often fraudulent ones). methods have been proposed to solve the imbalance problem, among most popular being oversampling techniques (such as SMOTE). These generate synthetic in class, balance dataset, performing data augmentations improve performance predictive (ML). In this paper, we advance novel, augmentation method (adapted eXplainable AI), generates synthetic, counterfactual class. Unlike other techniques, adaptively combines existing using actual feature-values rather interpolating values between instances. Several experiments four different classifiers and 25 involving binary classes are reported, which show Counterfactual Augmentation (CFA) useful datapoints The also CFA is competitive with methods, variants SMOTE. basis CFA’s discussed, along conditions under it likely perform better or worse future tests.