A novel tree-based dynamic heterogeneous ensemble method for credit scoring

作者：

Highlights：

• A tree-based heterogeneous ensemble credit scoring model is proposed.

• Advanced GBDT-based methods function as components of our proposal.

• An overfitting-cautious ensemble selection strategy is developed.

• Our proposal outperforms the benchmark models significantly in most cases.

• Our proposal is robust to slight modification on base model and fitness function.

摘要

•A tree-based heterogeneous ensemble credit scoring model is proposed.•Advanced GBDT-based methods function as components of our proposal.•An overfitting-cautious ensemble selection strategy is developed.•Our proposal outperforms the benchmark models significantly in most cases.•Our proposal is robust to slight modification on base model and fitness function.

论文关键词：Credit scoring,Selective ensemble,Random forests,Gradient boosting decision tree,Machine learning

论文评审过程：Received 9 September 2019, Revised 27 March 2020, Accepted 30 May 2020, Available online 7 June 2020, Version of Record 16 June 2020.

论文官网地址：https://doi.org/10.1016/j.eswa.2020.113615