Credit Risk Research Of Commercial Bank Based On Dynamic KMV Model And Timing Association Rules

Credit risk is the oldest and overriding one among the numerous risks confront by the commercial banks. Recently, with the deceleration in economic growth and adjustment in economic structure, the balance and ratio of non-performing loan keep increase constantly, which poses high requirement on credit risk management for China’s banking industry. However, the credit rating system and credit risk management in China are far behind that in western countries. The common macro factors such as economic condition and economic prospect confront by the load companies contributes to the increasing correlation of default. In addition, due to the cross-shareholding and interaction among different-ranking companies, a company’s default will implicate the others. Therefore, the evaluation on credit risk contagion effect among loan companies is an important part of credit risk management for commercial banks.In order to improve the bank’s credit risk management, the pivotal measure is to take a deep study on credit risk identification and measurement technology. This thesis firstly make a review on the previous research in credit risk measurement and credit risk contagion effect, furthermore, the mainstream credit risk assessment models are described and compared in detail. We figure out that the KMV model is the most applicable in our country. However, the current domestic studies on KMV model are on the basis of static assumptions to measure a company’s credit risk within a fixed period. In order to satisfy the commercial banks’ requirement on real-time monitor the credit risk condition of loan companies; to take the dynamic input parameters of KMV model into consideration, we modify the established KMV model by adding temporal dimension to set up a dynamic credit risk measurement model. Empirical study is also included in our thesis. We select 30 companies with a label of Special Treatment(ST) and 30 non-ST companies from 19 industries in 2014 as samples. In order to minimize the impact of unrelated factors on our study results, we restrict the paired non-ST companies in the same industry, stock exchanges with ST companies. Finally, the empirical results are tested by parametric and non-parametric hypothesis. It indicates that the dynamic KMV model established in this thesis can effectively distinguish the credit risk difference between ST companies and non-ST companies. In addition, compared with the theoretical expected default rates(EDF), the default distance(DD) is a more appropriate index for credit risk measurement in our country.Because of the large scale of credit assets managed by commercial banks, it is difficult to identify relationships between loan companies. Meanwhile, the traditional credit risk infection models are complex in theory, strict in assumption, intensive in calculation, which result in its difficulty in describing credit risk infection among companies. Therefore, this thesis applies the algorithm of association rules in data mining area into the study on credit risk infection to overcome the limitation of traditional models. This method takes advantage of data mining technology to figure out the association rules among companies and even dig out some hidden information. Traditional Apriori algorithm of association rules is one-dimensional while this thesis is two-dimensional because it is on the basis of dynamic credit risk measurement and equipped with temporal dimension. Combining meta-rule theory and time-window technology, we extend the Apriori algorithm to establish algorithm of temporal sequence association rules and dig into the association rules on daily default distance. We explain and proof the15 strong association rules yielded in the empirical study, and the results demonstrate that the strong association rules based on algorithm of temporal sequence association rules consonant with the objective fact. It can effectively figure out the associated relationship among loan companies.All in all, in order to capture real-time credit risk condition and to predict credit risk infection among companies, this thesis build up a dynamic frame to measure credit risk and deter infection. Constructing a dynamic KMV model can capture daily credit risk while establishing temporal sequence association rules can distinguish credit risk infection conditions among companies. The effectiveness of those models is testified by the empirical study. The dynamic model introduced in this thesis can help commercial banks adjust credit risk management strategy in time, control the scale of loans and evade risk effectively.