Research On Business Credit Evaluation And Risk Management Of Small And Micro-sized Enterprises

As an important part of China's economy,small and micro-sized enterprises have played a pivotal role in promoting economic growth and increasing employment opportunities.In recent years,the state has attached great importance to policies of financial inclusion,and has regarded“supporting the development of small and micro-sized enterprises and deepening financial services for small and micro-sized enterprises”as one of the key contents of the national strategy.However,since the information asymmetry between commercial banks and small and micro-sized enterprises is more serious than that of large enterprises,commercial banks are more difficult to judge the business credit risk level of small and micro-sized enterprises accurately under this conditions.In addition,since commercial banks have not established an effective“business credit evaluation and risk management model”specifically for small and micro-sized enterprises,the unit cost of business credit risk assessment for commercial banks is high and the predictive accuracy is low,which results in a higher default rate for small and micro-sized enterprises than that for large enterprises.In order to fundamentally solve the problems of small and micro-sized enterprises'financing difficulties and the commercial banks'high default rate caused by small and micro-sized enterprises'business credit risks,the key is to continuously improve the commercial banks'business credit management level and business credit risk control ability.In theory,the business credit risk of small and micro-sized enterprises is the result of the dynamic evolutionary game between banks and enterprises under asymmetric information.Therefore,to build an effective business credit risk mitigation mechanism so that the equilibrium tends to the evolutionary stable equilibrium,is the fundamental way to reduce the business credit risk of small and micro-sized enterprises in theory.Meanwhile,to establish an effective“small and micro-sized enterprises'business credit evaluation and risk management model”so as to improve the commercial banks'business credit risks assessment level is the effect way to solve the financing difficulties of small and micro-sized enterprises as well as to reduce the high default rate of commercial banks in practice.Under the above-mentioned theoretical background and realistic background,based on the problems of small and micro-sized enterprises'financing difficulties and high default rate of commercial banks,this dissertation conducts both corresponding theoretical analysis and empirical research.This dissertation attempts to establish a set of practical business credit evaluation and risk management models through the theoretical research on the formation mechanism and mitigation mechanism of small and micro-sized enterprises'business credit risk.The main research contents of the dissertation are as follows:Firstly,this dissertation conducts a theoretical research on the formation mechanism and mitigation mechanism of small and micro-sized enterprises'business credit risk.In this part,this dissertation first analyzes the general process of the formation and evolution of enterprises'business credit risk.And then,based on the evolutionary game model,this dissertation analyzes how the bank-enterprise gaming behavior strategy changes under asymmetric information,and further reveals the dynamics formation mechanism of business credit risk.Furthermore,based on the“Social Capital Theory”and“Transaction Cost Theory”,the business credit risk mitigation mechanism is established.Finally,based on the key issues existing in the business credit evaluation and risk management process(business credit judgment,business credit risk tracking,business credit risk early warning),this dissertation constructs a business credit evaluation and risk management control concept model,and thereby comes up with an empirical research path of small and micro-sized enterprises'business credit evaluation and risk management.The results show that:under information asymmetry condition,the groups of small and micro-sized enterprises and commercial banks in the loan transaction process will dynamically evolve towards the equilibrium point of non-cooperation.The results of the mitigation mechanism based on the“Social Capital Theory”lies in constructing the business credit evaluation indicators and business credit risk evaluation indicators of small and micro-sized enterprises(that is,the“business credit capital”acts as a“virtual collateral”to increase the trust between banks and enterprises before lending,and to establish reputation constraints for small and micro-sized enterprises after lending),which will ultimately promote the banks'group and enterprises'group dynamically evolve to the equilibrium stable strategy.The results of the mitigation mechanism based on the“Transaction Cost Theory”is to reduce the transaction costs of commercial banks and increase the transaction costs of small and micro-sized enterprises by improving the business credit evaluation and risk management models,which will ultimately promote the banks'group and enterprises'group dynamically evolve to the equilibrium stable strategy.Secondly,this dissertation conducts a research on business credit judgment of small and micro-sized enterprises.In this part,this dissertation regards“small and micro-sized enterprises applying for loans”as the research object and first analyzes the status quo in the management of business credit judgments of small and micro-sized enterprises.And then,based on the“Social Capital Theory”,a small and micro-sized enterprises'business credit evaluation indicator system is constructed from the perspective of the indicator system improvement,which mainly includes compliance indicators,practice indicators and public opinion indicators.Moreover,based on the“Transaction Cost Theory”,the KFS-MCLOC model is established from the perspective of model improvement,by introducing the 1-norm of the feature core weight vector and the core feature selection pair into the objective function of the MCLOC model.Finally,the prediction performance of the KFS-MCLOC model is verified through the small and micro-sized enterprises'data set,and the prediction performance of the existing representative models is comprehensively compared;meanwhile,the effectiveness of the key business credit indicators selected by the model is also analyzed and discussed.The results show that:by introducing business credit information into the original information can effectively improve the prediction performance of the evaluation model;the KFS-MCLOC model has better prediction accuracy and interpretability than that of other models;the KFS-MCLOC model screens out 18 key indicators,which can greatly affect the business credit of small and micro-sized enterprises.Thirdly,this dissertation conducts a research on the business credit risk tracking of small and micro-sized enterprises.In this part,this dissertation takes“small and micro-sized enterprises with loans”as the research object and first analyzes the status quo in the tracking and management of business credit risks of small and micro-sized enterprises,and then,this dissertation puts forward the key issues that need to be solved urgently in tracking the business credit risks of small and micro-sized enterprises.Furthermore,based on the“Social Capital Theory”,the small and micro-sized enterprises'business credit risk evaluation indicator system is constructed,which mainly includes compliance indicators,practice indicators and public opinion indicators.Moreover,based on the“Transaction Cost Theory”,the XGBoost-Probit-SRXG multi-class heterogeneous ensemble model is established.And finally,the enterprises'data set is used to verify the prediction performance of the XGBoost-Probit-SRXG model and makes a comprehensive comparison with the prediction performance of the existing representative models.Meanwhile,it also compares the effectiveness of the key business credit risk indicators selected based on the importance score of the XGBoost model.The results show that:by introducing business credit risk information into the original information can effectively improve the prediction performance of the evaluation model;the XGBoost-Probit-SRXG model shows better prediction performance than that of other multi-class models in terms of prediction accuracy;the feature selection model screens out 16 key indicators that affect the business credit risk of small and micro-sized enterprises.Fourthly,this dissertation conducts a research on business credit risks early warning for small and micro-sized enterprises.This dissertation first analyzes the current situation of early-warning management of business credit risks for small and micro-sized enterprises,and puts forward the key issues that need to be solved urgently in the early-warning management of business credit risks.And then,from the perspective of model improvement,empirical researches are carried out from two aspects:single business credit risk early warning and comprehensive business credit risk early warning.In terms of single risk early warning,taking the“normal and concerned small and micro-sized enterprises”as the research object,the L₁-Logit model is used to screen key single business credit risk indicators;then a homogeneous ensemble model based on KNN voting weight improvement is established to distinguish between“normal small and micro-sized enterprises”and“concerned small and micro-sized enterprises”.In terms of comprehensive risk early warning,taking the relatively higher risk“concerned small and micro-sized enterprises”as the research object,the GAET-Kmeans model is constructed to determine the comprehensive business credit risk level for“concerned small and micro-sized enterprises”.The results of the single business credit risk warning research show that:by introducing the business credit risk indicators can more effectively improve the prediction performance of evaluation model;compared with other single classifier models,the homogeneous ensemble model has higher predictive performance.The results of the comprehensive business credit risk warning show that:the GAET model determines the comprehensive business credit risk score of“concerned small and micro-sized enterprises”.Moreover,the comprehensive business credit risk level of“concerned small and micro-sized enterprises”is assessed through unsupervised learning algorithm K-means.Meanwhile,the“concerned small and micro-sized enterprises”have higher risk levels in terms of compliance risk,practice risk and public opinion risk,collateral guarantee risk,etc.The innovations of the dissertation mainly include the following aspects:Firstly,the business credit and business credit risk evaluation indicator system of small and micro-sized enterprises is improved.Since the difficulty in obtaining business credit data,the business credit information was rarely included in the credit evaluation indicator system of small and micro-sized enterprises in the past.With the advancement of data mining technology and the interconnection of data from various government departments,it is more possible to obtain data such as corporate behavior trajectory data,government regulatory data,etc.Therefore,it becomes a reality to establish a credit evaluation indicator system for small and micro-sized enterprises based on business behavior.This dissertation makes some explorations in the areas of compliance information,practice information and public opinion information,etc.,which perfects the small and micro-sized enterprises'business credit and business credit risk evaluation indicator system,and thereby provides a reference to improve the effectiveness of business credit evaluation and business credit risk assessment of small and micro-sized enterprises.Secondly,this dissertation systematically analyzes the dynamic action path of the formation mechanism and mitigation mechanism of small and micro-sized enterprises'business credit risk.This dissertation first classifies the formation mechanism of business credit risk of small and micro-sized enterprises,and then based on the information asymmetry theory,this dissertation analyzes the dynamic formation mechanism of the business credit risk of small and micro-sized enterprises in the bank-enterprise game through the evolutionary game model.From the perspective of intangible capital,the business credit risk mitigation mechanism is established based on the“Social Capital Theory”,and meanwhile,from the perspective of tangible capital,the business credit risk mitigation mechanism is established based on the“Transaction Cost Theory”.In addition,the business credit risk control model based on cybernetics is designed,where this control model can provide a theoretical conceptual framework for commercial banks to effectively control the business credit risk of small and micro-sized enterprises.Unlike most literatures that only use general economic analysis methods,this dissertation,which contains both general economic analysis and dynamic derivation of evolutionary game models,is more referable.Therefore,this dissertation can not only enrich the theoretical research,but also provide a theoretical basis for the empirical research.Thirdly,this dissertation effectively improves the business credit judgment model,business credit risk tracking model,and business credit risk early warning model of small and micro-sized enterprises.By constructing a business credit judgment model based on KFS-MCLOC,the problem that banks cannot accurately identify small and micro-sized enterprises with good business credit is solved;by constructing a heterogeneous ensemble model based on XGBoost-Probit-SRXG,the business credit risk tracking problem of small and micro-sized enterprises is solved;by constructing a model based on L₁-Logit and KNN voting weight improvement,the single business credit risk early warning problem is solved;by constructing an unsupervised learning model based on GAET-Kmeans,the comprehensive business credit risk early warning problem is solved.In addition,this dissertation uses the multi-source data of small and micro-sized enterprises to empirically test the effectiveness of the constructed models,and the results verify that these models have greater advantages in terms of prediction accuracy,practical applicability,and interpretability,which provides a reference for the application of these models in reality.