Hybrid probabilistic sampling with random subspace for imbalanced data learning

Abstract

Class imbalance is one of the challenging problems for machine learning in many real-world applications. Other issues, such as within-class imbalance and high dimensionality, can exacerbate the problem. We propose a method HPS-DRS that combines two ideas: Hybrid Probabilistic Sampling technique ensemble with Diverse Random Subspace to address these issues. HPS improves the performance of traditional re-sampling algorithms with the aid of probability function, since it is not sufficient to simply manipulate the class sizes for imbalanced data with complex distribution. Moreover, DRS ensemble employs the minimum overlapping mechanism to provide diversity and weighted voting, so as to improve the generalization performance. The experimental results demonstrate that our method is efficient for learning from imbalanced data and can achieve better results than state-of-the-art methods for imbalanced data.