Growth potential of machine learning in credit risk predicting of farmers in the industry 4.0 era

Abstract

This paper aims to design a model framework for farmer credit risk assessment based on machine learning. It reduces the degree of credit risk misjudgement caused by the weak correlation between evaluation indicators and default status and imbalanced data. Based on the empirical analysis of 8624 farmers' data from a commercial bank in China, the average rank of the OPSO‐GINI‐FS model designed from the feature dimension is 1.29, which is higher than that of the OPSO‐GINI‐IS model designed from the indicator dimension (1.57). This means that our model has a higher default risk identification ability than the traditional one. And the META‐SAMPLER method of processing imbalanced data is also promising. Moreover, we found the machine learning designed in this paper has a higher ability to identify farmers' loan default than the traditional econometric methods. These findings establish the potential of machine learning in credit risk identification from a micro perspective.