AUC4.5: AUC-Based C4.5 Decision Tree Algorithm for Imbalanced Data Classification

Abstract

This paper presents a modification of Quinlan's C4.5 algorithm for imbalanced data classification. While the C4.5 algorithm uses the difference in information entropy to determine the goodness of a split, the proposed method, which is named AUC4.5, examines the difference in the area under the ROC curve (AUC) of a split. It implies that our method attempts to maximize the AUC value of a trained decision tree in order to cope with class imbalance in data. An extensive experimental study was performed on 20 real datasets from the machine learning repository at the University of California at Irvine, Irvine. The proposed AUC4.5 algorithm showed better classification than both the standard and cost-sensitive C4.5 algorithms.