Multi-objective optimization-based adaptive class-specific cost extreme learning machine for imbalanced classification

Imbalanced classification is a challenging task in the fields of machine learning and data mining. Cost-sensitive can tackle this issue by considering different misclassification costs classes. Weighted extreme (W-ELM) takes cost-sensitive strategy to alleviate bias towards majority class achieve better performance. However, W-ELM may not optimal weights for samples from classes due adoption empirical costs. In order solve issue, multi-objective optimization-based adaptive class-specific cost (MOAC-ELM) presented paper. To be specific, initial are first assigned depending on information. Based that, representation minority could enhanced adding penalty factors. addition, optimization with respect factors formulated automatically determine costs, which multiple performance criteria constructed comprehensively rate generalization gap. Finally, ensemble implemented make decisions after optimization. Accordingly, proposed MOAC-ELM an method good robustness imbalanced problems. Comprehensive experiments have been performed several benchmark datasets real-world application dataset. The statistical results demonstrate that competitive