An Empirical Study Of The Listed Company Rating, Using The Classification Tree Algorithm

China's securities market has experienced 10 years of development, as the end of 2005,stock markets in Shenzhen a-share listed companies has reached 1,500. In the face of sucha huge market size, either as a regulator for the China Securities Regulatory Commissionand the Stock Exchange or to the investors, if they can use some methods of a scientificevaluation system for listed companies financial situation undoubtedly is the lifeline of asuccessful opportunity.Since the 1960s, credit risk analysis is widely accepted as the pattern recognition.Classification tree methodology is based on statistical theory, and computer technology toachieve the non-parameters recognition.It has the speed, high accuracy, user-friendlyadvantages than other statistical techniques such that increasing application.The purpose of this study is to explore classification trees technical feasibility for listedcompanies in China. Through experiments choose listed companies financial indicatorsapplicable to the algorithm, experiment with different rating methods and certificateclassification tree superiority. Provide a reference listed companies rating methods for thedomestic rating agencies.Because enterprise credibility problem finally is the financial enterprise failure, throughthe use of foreign credit rating agencies indices, from China listed company's financialindicators derived the corresponding indices.Firstly introduced a number of creditevaluation methodologies, and the principles and content of the algorithms. Throughexperiments contrast, expounded the advantages of classification trees.In order to build a model, the paper introduced the latest classification softwareweak.First introduced the history and functions of weak. Introduced the realization of someof the selection algorithm. Secondly, through various software features alternatives choosea feature selection method that have the best results.Through a one yuan analysis, and classification tree analysis Logistic identify methodsof experimental results contrast, a classification tree algorithm that compared with othermethods of superiority.The overall accuracy of the situation is not very high, studied a combination ofincreasing classification tree algorithms. And through experimental data demonstratesignificant portfolio enhancement algorithms can indeed improve the accuracy.