Cost Learning Network for Imbalanced Classification
Chun-Yi Tu, Hsuan-Tien Lin · 2020
This paper proposes a method to improve the performance of imbalanced classification via reinforcement learning and cost-sensitive learning. Since the cost information is usually unavailable for cost-sensitive learning, we incorporate reinforcement learning to optimize the specified metric by adjusting the cost-matrix for the underlying cost-sensitive classifiers. Our experiment results show that, with the learned cost-matrix, the cost-sensitive classifiers can achieve better performance on several benchmark imbalanced data sets.