Anovel HEOMGA Approach for Class Imbalance Problem in the Application of Customer Churn Prediction

Making class balance is essential when learning from highly skewed datasets; otherwise, a learner may classify all instances to negative class, resulting in high false-negative rate. As result, precise balancing strategy required. Many researchers have investigated imbalance using Machine Learning (ML) methods due their powerful generalization performance and interpreting capabilities, comparing with random sampling techniques, handle the problem of preprocessing phase facilitate process improve results learners. In this research, an effective method called HEOMGA presented by combining Heterogeneous Euclidean-Overlap Metric (HEOM) Genetic Algorithm (GA) for oversampling minority class. The HEOM employed define fitness function GA. To assess proposed method, three benchmark datasets UCI repository domain customer churn prediction are examined different ML learners evaluated metrics. experiment show effectiveness compared some popular oversample methods, such as SMOTE, ADASYN, G Gaussian methods. significantly outperformed other terms recall, mean, AUC Wilcoxon signed-rank test used.