基于ADASYN的跨境电商小微企业信用风险模型优化研究

Research on Credit Risk Model Optimization of Small and Micro Enterprises in Cross-border E-commerce Based on ADASYN

随着数字经济在我国的不断发展,数字经济技术为地方企业带来新的机遇与挑战。聚焦跨境电商小微企业信用风险管理,并以江西为例建立符合江西跨境电商实际发展境况的企业信用风险评价指标。随后分别使用BP神经网络、K近邻及支持向量机机器学习分类模型,对江西跨境电商小微企业进行信贷违约预测。实证结果显示:三种深度学习分类模型均具有良好的准确性。为进一步优化江西跨境电商小微企业信用风险模型,特引入ADASYN算法,研究结果显示:ADASYN算法优化了三种模型的预测准确度。将模型的计算效率用于衡量模型的优秀程度,发现ADASYN—KNN综合时效性最佳。

As the continuous development of digital economy in our country,digital economy and technology bring new opportunities and challenges for local businesses.This paper focuses on the credit risk management of small and micro enterprises in cross-border e-commerce,and takes Jiangxi as an example to establish an enterprise credit risk evaluation index that conforms to the actual development situation of cross-border e-commerce in Jiangxi.Then,BP neural network,K-nearest neighbor and SVM machine learning classification models are used respectively to forecast credit default of small and micro cross-border e-commerce enterprises in Jiangxi Province.The empirical results show that:the three deep learning classification models all have good accuracy.In order to further optimize the credit risk model of small and micro cross-border e-commerce enterprises in Jiangxi,ADASYN algorithm is introduced,studies show that ADASYN algorithm optimizes the prediction accuracy of the three models.Using the computational efficiency of the model to measure the excellence of the model,the author finds that the ADASYN-KNN comprehensive timeliness is the best.