| In recent years, as the main body of the social economy power,enterprises fell into the credit crisis more and morefrequently. Credit problems of commerce, products and internalmanagement have seriously affect the normal operation of China’seconomic order. The other participants of economic activities is also anurgent need for an effective and feasible means to distinguish differentkinds of enterprises.Enterprise credit rating provides an importantreference with its independent and objective scientific results to creditors,investors and even consumers. And the specific rating technology is akey factor in the whole rating activity.This article was from the summary of the existing ratingmethodology, finally took the neural network method as basic researchdirection. The primary coverage included: Firstly, this article analyzedcredit rating indicator system of main rating organizations and actualsituation of our country.Then established principle for rating targetselection and has completed the target selection. Secondly, the articlequoted the Adaboost-BP method to enterprise credit rating through itsapplicability was analyzed. Next, the empirical analysis processsuccessively carried on the pretreatment using the non-parameter methodand the principal components analysis to the sample data, then used the MATLAB software to complete the training process. And empiricalanalysis results showed that Adaboost-BP method was obviously higherthan the BP neural network method in the accuracy of discrimination,and significantly enhance the role of the original BP neural network.Atthe same time, as to the rating cost, the Adaboost-BP method can alsoapply well in actual. In addition, its simple describing pattern, fastsolution speed and higher tolerance of noise data makes it can be moreeffectively applied to the enterprise credit rating. Finally, this paperseparately from the enterprise, the enterprise credit rating organizationand investor’s angle put forward related proposal. |
PDF Full Text Request