Predicting mortgage default using convolutional neural networks

作者：

Highlights：

• A new model for credit scoring is introduced.

• The model relies on a convolutional neural net applied to historical account balances.

• The model is evaluated on a subset of DNB ASA’s Norwegian mortgage portfolio.

摘要

•A new model for credit scoring is introduced.•The model relies on a convolutional neural net applied to historical account balances.•The model is evaluated on a subset of DNB ASA’s Norwegian mortgage portfolio.

论文关键词：Consumer credit risk,Machine learning,Deep learning,Mortgage default model,Time series

论文评审过程：Received 15 August 2017, Revised 17 February 2018, Accepted 18 February 2018, Available online 19 February 2018, Version of Record 19 March 2018.

论文官网地址：https://doi.org/10.1016/j.eswa.2018.02.029