出口跨境电商信用风险评价方法优化研究——基于多种机器学习算法的比较分析  

Research on Optimisation of Credit Risk Evaluation Methods for Export Cross-border E-commerce——Comparative analysis based on multiple machine learning algorithms

作　　者：李剑锋[1] 战琦璇 LI Jianfeng;ZHAN Qixuan

机构地区：[1]中国计量大学

出　　处：《价格理论与实践》2024年第11期169-174,共6页Price:Theory & Practice

摘　　要：出口跨境电商在促进我国外贸提质增效、扩大外循环以及推动产业链迈向中高端等方面起到了重要作用。数字技术的创新进一步加剧了跨境电商的信用问题,亟须提高对出口跨境电商的信用风险评价能力。本文基于2012-2022年我国45家出口跨境电商沪深A股上市公司的317条数据,分别运用Logistic、SVM、MLP、Random Forest、XGBoost算法,建立出口跨境电商上市公司信用风险评价模型。结果显示:机器学习方法较传统回归方法更适合构建出口跨境电商信用风险模型;结合特征筛选的结果来看,SMOTE-Lasso CV-Random Forest是表现最好的模型。另外,相对重要性较高的指标依次为上市年限、标准审计意见、管理层平均年龄、独立站和资产负债率。基于此,应制定出口跨境电商信用风险评价标准,完善大数据技术以提高预测准确率;加快跨境独立站发展,推动跨境电商品牌建设;切实发挥出口信用保险作用,提升对贸易高质量发展服务质效。Export cross-border e-commerce has played an important role in promoting the quality and efficiency of China's foreign trade,expanding the outward circulation as well as promoting the industrial chain towards the mid-high end.The innovation of digital technology further aggravates the credit problem of cross-border e-commerce,and there is an urgent need to improve the credit risk evaluation capability of export cross-border e-commerce.This paper establishes a credit risk e-valuation model for listed companies of export cross-border e-commerce based on 317 data of 45 listed companies of export cross-border e-commerce in Shanghai and Shenzhen A-share companies in China from 2012 to 2022 by using Logistic,SVM,MLP,RandomForest,and XGBoost algorithms respectively.The results show that the machine learning method is more suit-able for constructing the credit risk model of export cross-border e-commerce than the traditional regression method;SMOTE-LassoCV-RandomForest is the best performing model combined with the results of feature screening.In addition,the indicators with higher relative importance are,in order,the number of years of listing,standard audit opinion,average age of management,independent station and asset-liability ratio.Based on this,credit risk evaluation standards for export cross-border e-commerce should be formulated,and big data technology should be improved to enhance prediction accuracy;the develop-ment of cross-border independent stations should be accelerated to promote cross-border e-commerce branding;and the role of export credit insurance should be effectively utilised to enhance the quality of services for the high-quality development of trade.

关 键 词：出口跨境电商 信用风险评价 不平衡数据 机器学习 

分 类 号：F83[经济管理—金融学]