On the combination of graph data for assessing thin-file borrowers’ creditworthiness

作者：

Highlights：

• We propose a new framework to combine graph representation learning methods.

• The proposed model overperforms the traditional methods to incorporate graph data.

• Graph data enhance creditworthiness-assessment performance.

• Graph data contribute the most to improving unbanked business credit assessment.

摘要

•We propose a new framework to combine graph representation learning methods.•The proposed model overperforms the traditional methods to incorporate graph data.•Graph data enhance creditworthiness-assessment performance.•Graph data contribute the most to improving unbanked business credit assessment.

论文关键词：Credit scoring,Machine learning,Social network analysis,Network data,Graph neural networks

论文评审过程：Received 27 January 2022, Revised 14 July 2022, Accepted 7 September 2022, Available online 16 September 2022, Version of Record 26 September 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.118809