Research On Credit Rating Of Private Enterprises

As the establishment and the progressive development of socialist market economic structure, the private economy in China have been resumed and developed, and it has become an indispensable component of the national economy. Nowadays, the development of the socialist market economic system is still not perfect, thus the absence of a social credit rating system affect the overall economic effect, which is a huge obstacle to the further development of private enterprises. Credit rating system consists of many elements, the paper focused on the method of the country's private enterprise's credit rating in terms of its basic preparation conditions and the testing of its model construction.I. Basic preparations of the method for construction of private enterprises'credit ratingFirst, in order to make a breakthrough on the methods, the paper reviewed the theoretical study and practical results in the field of the domestic and foreign enterprises'credit rating, and summed up the subjective experts discrimination theory (the traditional method) and credit ratings of statistical methods raised in the 50s and 60s of the 20th centuries respectively, with a view to innovate on the basis of the previous study.Secondly, to clarify the scope of this paper, the related concept of credit rating, the credit, business credit, credit risk, credit ratings and corporate credit rating and other related concepts connotation were illustrated, and the meaning of an enterprises'credit rating was expounded and Chinese enterprises'development and the status quo of the credit rating was summarized. More over, the same method applied to different subjects distinctively, Chinese private enterprises have certain characteristics of the times, their development obviously reflected policy-oriented role, so accurately grasping the meaning of Chinese private enterprises and development underlies a specific study. This paper induced on the development of Chinese private economy, and on the Chinese private enterprises'credit status from the aspects of lack of credit in policy, in market, and in welfare, and made a theoretical analysis on the absence of Chinese enterprises'credit by the cost—profit method, to have an overall understanding and grasping of the characteristics and the situation of the development of Chinese private enterprises'credit, so as to build good credit rating infrastructure. Finally, the indicator system is the basis for the credit rating, and an objective and fair means to ensure the results of credit rating. After studying the characteristics of private enterprises, to achieve the reasonableness of construction and operation of rating model, the paper made a theoretical analysis from setting principles, its connotation, and rating criteria of the private enterprises credit rating index system, in accordance with the request of safety, profitability, growth, transmission and productivity, in conjunction with the proceeding discussion, and it in principle, determined the viability of assets, profitability, solvency, ability to grow, cash flow and corporate management and the total asset turnover rate and so forth 10 key indicators, plus accounts receivable turnover rate and so forth 16 indicators as subsidiary, so as to provide guiding principles for the establishment of Chinese private enterprises credit rating index system, and to guarantee the accuracy of the results.II. Building and testing of Chinese private enterprises'credit rating model1. Building and testing of single modelMultivariate adaptive regression spline (MARS model or MARS) is a non-linear, non-parameter regression method. MARS model is adept in finding out the optimal interactive and replacement variable, easy to deal with high dimensional data hidden in the complex data structure, and can efficiently reveal important data forms and data relation, and its vantage points lie in the unnecessary of strong model assumption, data driven modeling process, ascertaining contribution of certain variables, better ability to explain model, faster computing and so on. Therefore this paper, based on the characteristics of Chinese private enterprises'credit, introduced the MARS methods to out country's credit rating, established private enterprises MARS credit rating model, and used 2003 private listed companies'financial databases in CCER Chinese economic and financial database and the database of the listing corporate management structure in 2003 to do empirical data testing. In the empirical testing, first use MARS model to select variables: the calculation of contribution of variables by MARS, ultimately settled the net assets of a share, operating profit growth rate, price-earnings ratio of a share, the main business income growth and the number of the independent directors. Then, test the above-mentioned five rating variables by the model, the test results indicate: accuracy rate for the business model of the default judgment is 54.17 percent, while that for the non-corporate default judgment is 96.6 percent; its average accuracy rate of two is 91.52 percent. Using the traditional logistic regression model and the non-statistical model RBF neural network method to the same database information, test results showed that the accurate rates of default judgment were 41.67 percent and 12.5 percent respectively, and those of non—default judgment of the enterprises are both 96 percent, with an average accuracy rate of 90.03 percent and 89.57 percent. Through comparison, the results of MARS credit rating on private enterprises, in view of the model accuracy, stability and misjudgment of the price performance, turned out better, especially non—business prediction accuracy rate was far higher than the RBF logistic regression model and network model. Therefore, this paper argues that if use a single model of Chinese private enterprises credit rating, MARS approach is applicable.2. Construction and testing of hybrid modelTo further improve the accuracy and stability of the credit rating of private enterprise model, we established two hybrid neural network models: hybrid model I, and hybrid model II.The building and testing of hybrid model I. By analysis of the application of single logistic regression methods, RBF and MARS network respectively to Chinese private enterprises credit rating to identify variables, MARS can settle the contribution of a variable and RBF network van fit the advantages of high accuracy; using mixture of MARS and RBF network methods to establish a hybrid neural network model, the variables automatically elected by MARS model as RBF neural network model's input to establish a two stage hybrid neural network. The test showed hybrid model I optimized the network structure, improved forecasting accuracy, also default judgment accuracy rate of hybrid model I was 50 percent, and non-default judgment enterprises accuracy rate is 100 percent, both average accuracy rate was 94 percent.Construction and testing of hybrid model II. Using some of the finding s of hybrid model I, combined the important variable selected by MARS model and the predictive result T of the output of the MARS model into RBF neural network modules, as a new variable can be seen as a priori knowledge to provide the tip to a neural network, the model greatly increase the accuracy and the stability. The results showed that, the default judgment accuracy of hybrid model II was 70.83 percent. Accuracy rate for the non-corporate default judgment was 100 percent; its average of the two was 96.5 percent.3. comparison between hybrid model and single modelConsidered accuracy, stability and misjudgment of the price integrate, compared MARS model( as the best rating model) and hybrid neural network model, we can see the accuracy and stability of the hybrid neural network model should be better than a single model, and hybrid model II can help achieve better rating results.III. Suggestions on the building of Chinese private enterprises credit rating modelIn light of the actual situation of Chinese private enterprises and the construction of credit rating system, in the situations that the conditions are satisfied, the hybrid neural network model would be the best choice for the private enterprises'credit rating to achieve accuracy and stability in the situation that the condition are unsatisfied, the sub—optimal choice is to adopt the MARS method to continue the single model rating.