A Comparative Study of Cost-Sensitive Boosting Algorithms

Abstract

This paper describes a study of different adaptations of boosting algorithms for cost-sensitive classification. The purpose of the study is to improve our understanding of the behavior of various cost-sensitive boosting algorithms and how variations in the boosting procedure affect misclassification cost and high cost error. We find that boosting can be simplified for cost-sensitive classification. A new variant, which excludes a factor used in ordinary boosting, performs best at minimizing high cost errors and it almost always performs better than AdaBoost. We also find that cost-sensitive boosting seeks to minimize high cost errors rather than cost, and a minimum expected cost criterion, applied during classification, greatly enhances the performance of all cost-sensitive adaptations of boosting algorithms. We show a strong correlation between an algorithm that produces small model size and its success in reducing high cost errors. For a recently proposed method, AdaCost,...