Whale Optimization Algorithm Combined with Oversampling Technique for Imbalanced Issues

Many sampling-based data preprocessing strategies are proposed to resolve imbalanced problems. The essential principle of sampling is to generate minority samples or filter majority samples based on a raw unbalanced dataset and finally get a rebalanced dataset. Combining oversampling methods with whale optimization algorithm for imbalanced issues will enhance the performance of existing oversampling algorithms. WOA-Oversampling consists of two main steps: reconcile train set with a specific oversampling technique and find the best set from the balanced train set using WOA. The biggest advantage of the algorithm is that it will make full use of existing oversampling methods and procure perfect train set. Experiments conducted on 10 imbalanced datasets compared with four existing overs ampling methods demonstrate the effectiveness of the algorithm.