Credit Scoring Model Based On LightGBM-BOA And Its Empirical Research

Small enterprises play an important role in the economic development and employment creation of our country.At present,the small and medium-sized enterprises account for more than 99% of the total number of enterprises in our country,absorbing 75% of the employment of the whole society.However,at present,the proportion of small enterprises to obtain bank credit is still low,and the strength of obtaining financial institutions' support is not commensurate with its position in economic development.The difficulty of loan acquiring has become a bottleneck restricting the development of small enterprises in China.To be exact,due to the imperfect financial information of small enterprises and the lack of necessary collateral and other congenital conditions,it is difficult for financial institutions to accurately measure the credit risk of small enterprises,and the problem of small enterprises' loan difficulty still exists.Therefore,it is necessary to build a credit scoring model with better resolution and accurately measure the credit risk of small enterprises with the help of artificial intelligence data mining technology,so as to provide theoretical support and decision-making reference for alleviating the current situation of small enterprises' loan difficulty.Using machine learning technology,this paper presents a LightGBM credit scoring model based on Bayesian optimization,and conducts empirical research on 1820 industrial small enterprise data sets in China,690 in Australia and 1000 loan sample sets in Germany.On the small enterprise data set,LightGBM-BOA credit scoring model is established and the default status of small enterprises is predicted.By calculating the index importance score,by calculating the importance score,this paper selects 21 key indicators,such as quick ratio,sales net interest rate,per capita disposable income of urban residents,which have a great impact on the credit risk of small industrial enterprises,and establishes a credit rating indicator system of industrial small enterprises.Finally,according to the LightGBM model,the default probability of the loan enterprise is predicted,and the credit rating of the loan enterprise is divided according to the calculated score.The structure of the article is divided into five parts,the contents are as follows.The first chapter is the introduction,which mainly introduces the research background and significance of credit scoring,and systematically summarizes the relevant literature.The second chapter is the concept definition and theoretical basis,which defines the concepts used in this paper,including the definition of small business,credit,credit rating,credit rating and Ensemble model.Then it introduces three related theories used in this paper,including the theoretical principle and calculation formula of LightGBM,the basic principle and solution steps of Bayesian optimization theory,the basic principle and calculation formula of credit scorecard.In the third chapter,a credit scoring model based on LightGBM-BOA algorithm is established.First,the overall framework of the credit scoring model is given,and the evaluation index of the model is determined.Finally,the modeling steps of the credit scoring model are given step by step.The fourth chapter is an empirical study.Lightgbm-boa scoring model is established on the data set of industrial small enterprises and the default state is predicted.Then,the attribute reduction,scorecard establishment and credit rating of industrial small enterprises are carried out.Finally,the robustness test is carried out on the data sets of Germany and Australia.The fifth chapter is conclusion and prospect.The empirical conclusions of this paper are summarized,and the shortcomings and prospects of model selection and data selection are put forward.The results show that:(1)the LightGBM-BOA model constructed in this paper is compared with 7 models,such as LR,LDA,KNN,CART,NB,RF and SVM.The empirical results of three datasets show that the LightGBM-BOA model proposed in this paper is better in the criteria of Accuracy,KS value,AUC compared to other models.(2)Through the analysis of the credit rating index system of industrial small enterprises constructed in this paper,it is found that the impact of financial factors on the credit risk of industrial small enterprises accounts for 40.53%,and the external macro conditions of the location of small industrial enterprises have a significant impact on their credit risk.Furthermore,because the importance score of single index of mortgage and pledge guarantee is the highest,it shows that mortgage and pledge guarantee is very important to the credit risk of industrial small enterprises.(3)This paper uses LightGBM-BOA model to test the data sets of Australia and Germany,and its AUC score is higher than the reported results in the existing literature,which shows that the LightGBM-BOA model proposed in this paper has high default prediction ability and robustness.