Handling imbalanced datasets:over fitting problem

One of the most persistent problems that are in the real-world application is the class imbalance problem (CIP). When the class distributions are highly imbalanced data set are called imbalanced data set. When the target class for traditional classifiers is smaller than other class, CIP occurs. By CIP, the training data will significantly influence the classification accuracy. Ensemble methods designed by researchers to avoid CIP. Through analyzing the overall accuracy of ensemble methods, results suffer from over fitting problem. This over fitting problem impressed other significant results of ensemble methods in classification of imbalanced data sets. This paper analyzes and show how over fitting problem can suffer ensemble methods’ results in overall accuracy. Consequently, verifies overcoming of over fitting problem in face of imbalanced data sets.

Keywords:

Imbalanced data , Over fitting , Ensemble Methods , class imbalance problem