Research On Enterprise Credit Evaluation Based On Support Vector Machine

In contemporary society,credit consumption is growing rapidly,not only personal credit loans,but also many enterprises.Therefore,in order to meet the needs of economic development,it is necessary to conduct a comprehensive credit evaluation on the consumption ability,social performance ability and reputation of individuals and even enterprises.The development of credit evaluation has gone through three stages: expert analysis,statistical analysis and artificial intelligence.However,due to the late construction of China's enterprise credit system,enterprise credit management means and evaluation methods can not meet the needs of current economic development,which seriously restricts the development of enterprise credit scale and consumption ability.Due to the single perspective of enterprise credit environment,with the rapid development of enterprise credit evaluation,it is no longer applicable.It is of great theoretical value and practical significance to explore scientific and reasonable enterprise credit evaluation methods for promoting the construction of China's credit system and social and economic development.Therefore,some institutions and scholars began to study industry competition,management,policy and other factors.And improve a variety of statistics,artificial intelligence and other modern technology,will be applied to credit assessment,in order to build a credit risk assessment system,the combination of historical experience and scientific evidence;coupled with the development of information technology and the progress of science and technology,slowly developed a series of credit assessment methods and models.At present,scientific credit evaluation method as a risk assessment tool has been recognized by the society and entrusted with important tasks.The most common application of statistical analysis knowledge has established many credit evaluation models,such as discriminant analysis(DA),logistic regression(LR)and classification tree(CT).Although using these models for credit evaluation,not only the implementation cost is low and relatively fast,but if the linear relationship between variables is lack,the accuracy of these models is not high.With the progress of the times and the development of technology,some artificial intelligence methods such as decision tree(DT),support vector machine(SVM),artificial neural network(ANNs),Bayesian classifier(BC),fuzzy rule system and integrated learning model have been used to construct accurate and stable credit risk assessment system.Therefore,this paper first expounds the definition and characteristics of credit risk,and then reviews the credit risk assessment and development process at home and abroad.Then it analyzes the credit evaluation factors at home and abroad,discusses different credit evaluation models,and analyzes and summarizes the commonly used mathematical statistics and artificial intelligence methods in enterprise credit evaluation.Through the comparative analysis of different credit evaluation models,in view of the current situation of enterprise credit risk evaluation in China,this paper selects SVM as the model basis,makes an in-depth study on the credit index system and SVM theory of domestic and foreign enterprises,and discusses the SVM theory.After that,the first mock exam of support vector machine kernel function combination and the combined evaluation model of SVM using AdaBoost integration algorithm are discussed.The data set of Poland bankruptcy company is used in the experiment.The result shows that the accuracy and practicability of the combination of kernel function or algorithm combination model are higher than that of single model.Then,in order to verify the superiority of the combination model again,the paper designs and verifies the enterprise credit evaluation model based on the combination of support vector machine,classification and regression tree under incremental learning,and uses the corresponding parts of the model to build a total IL-SVM-CART classifier model in different situations,and verifies its performance on the credit data set of South Germany in order to obtain higher credit rating Finally,the paper compares the effectiveness of the proposed method with other traditional methods in the evaluation of enterprise credit rating.