Would two-stage scoring models alleviate bank exposure to bad debt?

作者：

Highlights：

• We freshly use a sample of personal loans, provided by one of the largest Indian banks.

• We uniquely use actual misclassification costs to evaluate our models.

• Our scoring models can significantly reduce the default rate by 14.24%.

• Redundancy and branch-closure policies could thus ultimately have been avoided.

摘要

•We freshly use a sample of personal loans, provided by one of the largest Indian banks.•We uniquely use actual misclassification costs to evaluate our models.•Our scoring models can significantly reduce the default rate by 14.24%.•Redundancy and branch-closure policies could thus ultimately have been avoided.

论文关键词：Credit,Indian banks,Neural networks,Actual misclassification costs,Timing of Default

论文评审过程：Received 26 July 2018, Revised 14 March 2019, Accepted 15 March 2019, Available online 15 March 2019, Version of Record 21 March 2019.

论文官网地址：https://doi.org/10.1016/j.eswa.2019.03.028