New Study by Yoon Lee Tackles Imbalanced Data Sets in Business Analytics

Imbalanced data sets are a growing problem in data mining and business analytics. For example, the ability of machine learning algorithms to predict the minority class deteriorates in the presence of class imbalance. Although there have been many approaches that have been studied in literature to tackle the imbalance problem, most of these approaches have been met with limited success. A new study by Turner College management information systems professor Yoon Lee and Riyaz Sikora of the University of Texas at Arlington, which is scheduled for publication in a future issue of Information Systems Frontiers, proposes three methods based on a wrapper approach (i.e., a feature selection technique that finds the best subset of features for a specific machine learning model and domain) that combines the use of under-sampling with ensemble learning to improve the performance of standard data mining algorithms. Sikora and Lee test their ensemble methods on 10 data sets collected from the UCI repository with an imbalance ratio of at least 70%. They also compare the performance of their ensemble method to two other traditional techniques for dealing with the imbalance problem and show significant improvement in both recall and the average of precision and recall.