Benchmarking Swarm Rebalancing Algorithm for Relieving Imbalanced Machine Learning Problems

Imbalanced classification is a well-known NP-hard problem in data mining. Since there are more data from the majority classes than the minorities in imbalanced dataset, the resultant classifier would become over-fitted to the former and under-fitted to the latter. Previous solutions focus on increasing the learning sensitivity to the minorities and/or rebalancing sample sizes before learning. Using swarm intelligence algorithm, we propose a series of unified pre-processing approaches to address imbalanced classification problem. These methods used stochastic swarm heuristics to cooperatively optimize and fuse the distribution of an imbalanced training dataset. Foremost, as shown in our published paper, this series of algorithms indeed have an edge in relieving imbalanced problem. In this book chapter we take an in-depth and thorough evaluation of the performances of the contemporary swarm rebalancing algorithms. Through the experimental results, we observe that the proposed algorithms overcome the current 17 comparative algorithms. Though some are better than the others, in general these algorithm exhibit superior computational speed, high accuracy and acceptable reliability of classification model.