Credit Risk Assessment of Green Supply Chain Finance for Logistics Enterprises: A Comparative Study of Logistic Model and BP Neural Network

Abstract

With the escalating global emphasis on sustainable development, Green Supply Chain Finance (GSCF) has emerged as a critical area of interest for academic researchers, industry practitioners, and policy makers alike. However, logistics companies, as significant contributors to greenhouse gas emissions within supply chains, have been underrepresented in credit risk assessment research. This paper bridges this research gap by constructing a comprehensive credit risk evaluation framework encompassing four critical dimensions: financial status, asset status, green development, and supply chain status. The system is evaluated using a sample of 131 listed logistics enterprises from 2016 to 2021. After applying principal component analysis for dimension reduction, both a Logistic regression model and a BP neural network model are trained. The BP neural network model possesses exceptional nonlinear mapping capabilities, effectively capturing complex interactions among variables and overcoming the limitations of traditional linear regression models. The empirical results demonstrate that green development and supply chain sustainability indicators significantly predict enterprise compliance rates across both modeling approaches. Furthermore, key metrics demonstrate that the BP neural network model exhibits superior predictive accuracy and reduced Type II error propensity relative to the Logistic regression model, thereby presenting a more robust analytical tool for financial institutions. This paper contributes to the advancement of GSCF by providing actionable insights for logistics enterprises, financial institutions, and policy makers to develop a more supportive and robust GSCF system.