An Assesment Of Credit Risk Of Listed Companies Based On Data Mining&MCDM

As one most important unit of modern economy, the financial risk of listedcompanies has a significant impact on the national economy status. Thus, how torapidly and accurately predict the credit default of listed companies has become a veryimportant issue in the field of financial risk management.With the development of many disciplines, especially the database technology andcomputing power, the technology of data mining has been seen as a great integration ofmany traditional and new disciplines. It has a relative whole system of methods for datacleaning, feature selection, classification and prediction, etc., so it is very promising inthe field of credit risk assessment. The technology of multiple criteria decision making(MCDM) is quite suitable for handing problems with multiple inputs and outputs, andcan analyze the relationship and relative importance between attributes. Moreover, thecharacteristics of contradiction and non-unique standandard between multiple criteria ofMCDM coincide with the feature of financial sector, namely minimization ofinvestment risk with maximization of profit, and this provides a possibility of theapplication of relative methods. In fact, the MCDM methods have been widely used inthe fields of credit risk assessment, investment projects ranking, and so on.Considering the complex characteristics of credit default risk of listed companies,the application of an integrated method of data mining and MCDM in default risksoring and default classification and prediction is of a considerable significance. Basedon data mining and MCDM, this research integrates the results of emergingcross-disciplines into the field of financial risk assessment and default prediction, andconstructs a related theoretic model, hoping to provide a new solution to the recognition,assessment and management of listed companies’ credit risk.This paper takes the credit risk status of over one hundred listed companies inShanghai and Shenzhen stock market as the object, systematically investigates themethods of data mining and MCDM for credit default prediction. Firstly, it uses severaltypical methods of data mining to preprocess the relative financial data and select keyattributes. Secondly, it completes the tasks of determination of index weights and ranking of decision-making units by means of typical MCDM methods. Lastly, it usesthe methods of clustering in data mining to predict the default status of companies, so asto differentiate between good and bad objects. In the aspect of the use of financial data,this paper adopts a different approach with previous research, which considers thetemporal factor to construct a dynamic assesment model including the time factor, andthus more in line with the characteristics of risk accumulation and continuous evolution.Experimental results show that the assessment and prediction model built in thisresearch has a good performance in credit risk ranking and default prediction for listedcompanies, with a relatively high accuracy. In addition, by using the different data setsto test the model, it further proves the applicability and scalability of the model.