A new technique for classification method with imbalanced training data

Imbalanced classification is a very common and crucial challenge in the machine learning domain. Due to unequal instances in different classes, the performance of traditional classifiers may decrease. How to deal with this imbalanced data is a major focus for the majority of studies. In terms of balancing the data, at the data level point under-sampling, over-sampling, and their variants are widely used. Since oversampling creates precise replicas of examples from the minority class, it may increase the risk that overfitting may occur. Undersampling wipes out a significant quantity of data, making it more difficult to determine where the decision boundary between minority and majority classes lies. In this work, we have proposed a novel method that combines both of these strategies based on the association rule and GAN to avoid this kind of problem score and produce a well-balanced data set. Extensive experiments have been conducted by using diverse benchmark datasets to compare this proposed approach with various state-of-the-art methods. The results demonstrate that the proposed approach outperforms existing techniques, achieving better classification performance while efficiently addressing the imbalanced classification problem.