Credit Rating Of Small And Medium Enterprises Based On BP Neural Network

The development of small and medium enterprises is related to the development of national economy and society.the vast majority of Chinese enterprises are small and medium enterprises,which play an important role in the employment of urban population,export and technological innovation.At present,commercial banks are the main source of financing.but the commercial banks worry about financing capability of small and medium enterprises.It has some reasons.on the one hand,the restriction of credit rating system and methods.Commercial banks can not assess very well the credit of small and medium enterprises.It leads to commercial banks not to loan.Therefore,improving the credit rating system of SME(small and medium enterprises)has become particularly important.The credit rating of enterprise is useful to commercial banks in assessing the credit risk of small and medium-sized enterprises,and help excellent small and medium-sized enterprise to accept more bank policy,funding tilt;at the same time,it plays an important role in warning to urge action of enterprise management,which have not well management.In the selection of the index system,combining with the qualitative and quantitative indicators,which provides the certain reference in complete SME credit rating system.The BP neural network is used for credit rating of small and medium sized enterprises.It enriched and updated the credit rating method of commercial bank.This has the important academic value.This paper put the financing problem of small and medium-sized enterprises as the starting point.Letting financing issues and credit rating of good docking,which provide a breakthrough to solve the financing problem of small and medium-sized enterprises.The credit rating index system of small and medium-sized enterprises have a combination of qualitative and quantitative indicators,and it quantify the qualitative index,which reduce the subjective factors in selection index.Scoring by BP neural network method,taking 50 enterprises in small and medium enterprise board as the sample,we score the credit,and also to test established a neural network model.At the same time,the author use the linear regression method to fit the sample.Then by comparing the results of the two models and analysis of their results,they further expounds the significant advantages of neural network in the field of the credit rating.At first,This article carry on the elaboration to the reasons of financing difficulties faced by small and medium-sized enterprises.The bank worries about credit of enterprise,which is one of the reasons,which led to this phenomenon.A breakthrough of solving the financing difficulties is putted forward to,which is to improve the small and medium enterprises credit rating system of commercial banks.Then,the paper refer to credit rating index of the people's Bank of China,The Association,China Construction Bank,standard&Poor's and Moody's.The small and medium-sized enterprises of our country in information is disclosed relatively small.There are some the financial index data that is not true.Combining with characteristics of this text,this paper established credit rating index system of small and medium-sized enterprises.six senior level indexes:debt paying ability,profit ability,operation ability,growth ability,the quality of the staff and manager,innovation ability,16 second grade indicators.In the selection of the second grade indicators,the author collect and arrange previous research findings.The intergrated analysis shows that 16 indexes.Index system combines the qualitative and quantitative indicators,and have the financial and non-financial indicators.Three qualitative indicators are management level of the leader,the quality of the staff and innovation ability,which are quantified.They are expressed by the administrative personnel proportion,college personnel proportion,scientific and technical personnel proportion which can quantify.The index system established is suitable for small and medium-sized enterprises,which provide a certain reference for the bank to assess corporate credit.After the construction of index system is completed,it is necessary to select a scoring model.5C,5Pand 5W is the traditional credit rating method.these methods for the factors that cannot be quantified with subjective uncertainty.And it is a high demand for personnel quantity and quality of experts.At the same time,it has to update the database of expert constantly.Application of statistical model in credit rating,which is simple and easy to operation.But the model does not reflect a dynamic process.Analytic hierarchy process(AHP)in credit rating,when it determines the relative importance,this has large subjectivity.However,in the process of the BP neural network theory,it can avoid the subjectivity in weight determination,also can be very well to deal with the nonlinear problems,have a very strong learning ability,and is suitable for credit rating of small and medium sized enterprises.After the selection of the above indexes and models,the empirical analysis based on the model is carried out.And the results are compared with the one of the linear regression model.According to the needs and the theory of neural network,the author build the BP neural network of 16-8-1 topological structure.It sets only one hidden layer of network structure,which the input layer has 16 nodes(16 secondary indexes),hidden layer with 8 neurons and an output value that is credit score of the enterprise.Taking the sample data of 50 small and medium enterprises as an example,using the sample partition method,40 samples are used to train the BP neural network,and the 10 samples are used to test the network structure.With an absolute error of less than 0.05 as the tolerance range,the accuracy of the training samples is as high as 92.5%,and the accuracy of the test sample is 80%.At the same time,using the same data,the paper do multiple linear regression on the above samples,dependent variables is expected credit score,the independent variable is 16 node indexes.Residuals of the fitting results is much larger.And the accuracy of sample fitting is only 42.5%,which show that linear regression in the credit rating of small and medium-sized enterprise has huge drawbacks.The empirical results show that the BP neural network has a great advantage in credit rating of the small and medium enterprises.Finally,the conclusions of this paper:(1)The asymmetric information is the main reason difficult financing problem of the small and medium-sized enterprise.In order to avoid risks,commercial banks do not loan to small and medium enterprises.Therefore,to strengthen information disclosure of the enterprise and the credit rating system of small and medium sized enterprises,which crack power of difficult financing problem.(2)Credit indicators of small and medium enterprises should be combined with qualitative indicators and quantitative indicators,financial indicators and non-financial indicators in the selection of indicators.And qualitative indicators are quantified.It will reduce the subjectivity of the indexes.(3)Credit rating method of statistical model is subjected to variable data which follow normal distribution.But financial data is generally not follow normal distribution.In the empirical comparison,it shows that the statistical model is not applicable to credit rating of small and medium sized enterprises.(4)At the same time,in the empirical research,also showed that the BP neural network have many advantages in the credit rating:Firstly,it has a good adaptive ability,in determining the weights of the index system and it does not need artificially to determine.According to the repeated training and learning of dataes,it adjust transmission and the relationship between input and output,which weaken subjective factors of the existence.Secondly,the ability to deal with the nonlinear problem.It gets the more residual in linear regression model,and relatively smaller residual in BP neural network,which reflects the ability to deal with nonlinear problems.Thirdly,the BP network has a very good dynamic evaluation results.