A cost-sensitive decision tree approach for fraud detection

作者：

Highlights：

• Cost-sensitive tree algorithm for credit card fraud detection with varying costs.

• Each fraud is assumed to have a different misclassification cost.

• Our approach can prioritize frauds according to their misclassification costs.

• A new cost-sensitive domain specific performance metric is defined.

• Cost-sensitive models show the best performance on tests with real world data.

摘要

•Cost-sensitive tree algorithm for credit card fraud detection with varying costs.•Each fraud is assumed to have a different misclassification cost.•Our approach can prioritize frauds according to their misclassification costs.•A new cost-sensitive domain specific performance metric is defined.•Cost-sensitive models show the best performance on tests with real world data.

论文关键词：Cost-sensitive modeling,Credit card fraud detection,Decision tree induction,Classification,Variable misclassification cost

论文评审过程：Available online 18 May 2013.

论文官网地址：https://doi.org/10.1016/j.eswa.2013.05.021