A Study On Small And Medium-sized Enterprises' Credit Risk Assessment In Supply Chain Finance

The solutions of small-and medium-sized enterprises(SMEs)' financing issues today are drawn the attention of academic and practical circles.Especially in China,SMEs are confronted with huge challenges such as capital constrained,high operational risks and financial information opaque.The essential problem is that SMEs are still squeezed out of conventional financing patterns even though the polices of the Chinese government are quite helpful.Based on theory and practice,the solutions of SME's financing issues today focus on the design and optimization of the finance pattern.As a means of substituting for lower credit availability,supply chain financing(SCF)is increasingly generating much enthusiasm of schorlars for solving the SME's financing issues.Nevertheless,SCF is unable to completely eliminate the SMEs'credit risk which is still one of the major threats to bank,which has a negative impact on SME's financing.It is well know that a fraction of improvement in assessment level also significantly promotes SCF in practice.The research of the paper is organized by three parts:(i)analysis of formation mechanism of SMEs' credit risk;(ii)construction of evaluation indexes system of SMEs' credit risk;(iii)construction of evaluation moldes of SMEs' credit risk and empirical study.In the part of analysis of formation mechanism of credit risk,we research the main participators in SCF,i.e.,SMEs,core enterprise(CE)and bank,and analyze the relationships of these participators at first.Second,we introduce three primary financing patterns in SCF:account receivable financing pattern in SCF,advance payment financing pattern in SCF and inventory financing pattern in SCF.Third,we discuss the influencing factors and embodies of SMEs' credit risk in SCF.Finally,we research the SMEs' credit risk in SCF by game theory method.In the part of construction of evaluation indexes system of credit risk,we firstly introuduce the current SMEs' credit risk of evaluation indexes system.Then,we design a new SMEs' credit risk of evaluation indexes system,which includes the design methods,design idea and basic framework.Finally,we construct a new SMEs' credit risk of evaluation indexes system which contains four main indexes,i.e.,SMEs of main index,CE of main index,assets base on financing of main index,supply chain operation of main index,and 18 sub indexes.Meanwhile,these 18 sub indexes are divided five classes,i.e.,liquidity,leverage,profitability,activity and non-finance.In the part of constructing evaluation moldes of SMEs' credit risk and empirical study,we evaluate the SMEs' credit risk in SCF basic on Logistic regression(LR)method,neural network(NN)methods,and machine learning(ML)methods.The models construction and empirical study primarily include three factors:(?)definitions of dependent variables and independent variables and constructions of each evaluation models,i.e.,LR model,radial basis function(RBF)NN,Logistic-RBF two stages hybrid model ?,?,?,individual ML model(decision tree,DT),ensemble ML models(i.e.,Bagging,AdaBoost,Random Subspace,Real AdaBoost),integrated ensemble ML models(Multi-Boosting,RS-Boosting,RS-RAB);(?)sampling procedure and data normalization;(?)the analysis of experimental results which include the selections of evaluation criteria and comparisons of experimental results.The evaluation criteria contains average predictive accuracy rate,type ? error,type ? error,recall rate,precision rate,F-Measure rate,value of receiver operating characteristic curve(ROC).The experimental results show that as follows:(?)the evaluation performance of RBF NN model is little better than that of LR model;(?)the evaluation performances of Logistic-RBF two stages hybrid models are obviously better than that of LR model and RBF NN model.Meanwhile,Logistic-RBF two stages hybrid model ? has the best performance among three hybrid models;(?)the evaluation performances of ensemble ML models are not always better than that of individual ML model;(?)the evaluation performances of integrated ensemble ML models are better than that of individual ML model and ensemble ML models.In specially,the RS-RAB gains the best performance in these ML models.This paper enriches the theory of SMEs' credit risk assessment under the background of SCF and provides significant references for assessing the SMEs' credit risk in practical work.