Adaptive multi-objective swarm fusion for imbalanced data classification.

•AMSCO gradually rebalances the imbalanced dataset through swarm optimizations.•Two parallel swarms separately sampling the two classes to decrease the search area.•The gradually rebalanced dataset has the maximized accuracy and reliability.•The rebalanced dataset size is controlled to preserve the original dataset size.•AMSCO exceeds several typical rebalancing techniques in all estimate metrics.