Handling Imbalanced Data Classification Problem using Artificial Immune System with Mahalanobis Distance

Most of the data in the classification domain are imbalanced data, meaning that the answers of each class are not the same. The data prediction in the majority class has been accurate while the answer that is interested is in the data in the minor class. To overcome this problem, we propose a sampling technique using the artificial immune system with Mahalanobis distance (AIS/MD) for the imbalanced class distribution. The proposed method is to reduce the complexity of the original process used by the artificial immune system class, which has proven to be beneficial and promising in different applications. Experimental results on 10 typical imbalanced datasets from UCI Machine Learning Repository show that the proposed method achieves better general implications compared to other methods.