Demonstrating non-inferiority of easy interpretable methods for insolvency prediction

作者：

Highlights：

• Insolvency prediction using a database with about 5000 enterprises.

• Comparison of white-box models with gray- and black-box models.

• Comparison of interpretability of disjunctive normal forms and decision trees.

• Use of asymmetric bagging to deal with imbalanced data.

• Interpretable models are not inferior to black-box models in insolvency prediction.

摘要

•Insolvency prediction using a database with about 5000 enterprises.•Comparison of white-box models with gray- and black-box models.•Comparison of interpretability of disjunctive normal forms and decision trees.•Use of asymmetric bagging to deal with imbalanced data.•Interpretable models are not inferior to black-box models in insolvency prediction.

论文关键词：Insolvency prediction,Interpretable model,Decision trees,Disjunctive normal forms

论文评审过程：Received 10 October 2014, Revised 5 August 2015, Accepted 5 August 2015, Available online 14 August 2015, Version of Record 3 September 2015.

论文官网地址：https://doi.org/10.1016/j.eswa.2015.08.009