Online Learning With Adaptive Rebalancing in Nonstationary Environments

An enormous and ever-growing volume of data is nowadays becoming available in a sequential fashion various real-world applications. Learning nonstationary environments constitutes major challenge, this problem becomes orders magnitude more complex the presence class imbalance. We provide new insights into learning from imbalanced online learning, largely unexplored area. propose novel Adaptive REBAlancing (AREBA) algorithm that selectively includes training set subset majority minority examples appeared so far, while at its heart lies an adaptive mechanism to continually maintain balance between selected examples. compare AREBA with strong baselines other state-of-the-art algorithms perform extensive experimental work scenarios imbalance rates different concept drift types on both synthetic data. significantly outperforms rest respect speed quality. Our code made publicly scientific community.