Credit Risk Evaluation Model of Supply Chain Finance Based on Deep Dimension Reduction

【 Objective: 】 A credit risk assessment model is named SAE_DE_SVM based on deep learning dimension reduction is proposed to solve the problems of multi-heterogeneous and dynamic high-dimensional characteristics in credit risk assessment of supply chain finance. 【 Methods: 】 This study obtains samples from CSMAR database, Sina Finance and Economics Network and Shenzhen Stock Exchange official website, and uses Stacked Auto-Encoder (SAE) to reduce the dimension of supply chain financial risk assessment features. Considering the imbalance between the positive and negative proportions of the evaluation samples, the Synthetic Minority Oversampling Technique (SMOTE) oversampling technique is used to balance the samples. Finally, the differential evolution (DE) algorithm is used to optimize the support vector machine (SVM), and SAE_DE_SVM algorithm as supply chain financial credit risk evaluation model is constructed. 【Results】 The results show that the accuracy and time complexity of SAE_DE_SVM model on supply chain financial sample data are 95.83 % and 5.56 s, respectively, which is the best in the comparison model. 【 Limitations】 In the process of deep learning dimension reduction, a part of the feature data and information will be lost. However, the related research on the accurate calculation and utilization of these data and information loss is still very lacking. 【Conclusion】 The experimental results show that credit risk assessment model of supply chain finance based on SAE_DE_SVM has good performance in predicting the possibility of default of Small and Medium-sized Enterprises (SMEs).