The Comparison Of Logistic Regression And Discriminant Analysis In Commercial Bank Credit Rating

Since the late 1990s, credit risk management has become the most crucial and challenging factors of risk management. Commercial banks bear the credit risk management, so how to prevent and resolve the credit risk has become issue. With China's accession to WTO, China has played the more and more important role in the international economic and financial cycle system. Only if commercial banks in China establish the credit risk system as soon as possible, will they enhance competitiveness and achieve sustained, healthy and stable development.Credit rating is the cornerstone of the credit risk management. Credit rat-ing is a way to measure the borrower's PD and EAD in order to offer a loan basis for decision-making and risk management. Credit rating is the basis of risk measurement, no credit rating no real risk management.In this paper, Fisher discriminant analysis and Logistic regression are used to establish the probability of default model. The paper introduce the concept of entropy and give the amount of information presentation, using the amount of information to select the size of financial indicators.In the end, combined with the real data of a commercial bank to estimate the PD. ROC and CAP curves and Kolmogorov-Smirnov test are used to compare the stability and efficacy between Fisher discriminant analysis and Logisitc regression.