A Study On Commercial Bank Credit Risk Early-warning Management System

The credit risk, comprising liquidity, investment and loan risk, is what faces commercial banks. So it is necessary to work out the early-warning system against credit risk for the commercial banks on the basis of modern system theory, of finance management theory, and of the experience obtained from financial supervision.Risk identification is the precondition of building the early-warning system. Nonlinear mechanism of bank system, which may be depicted from six sides, has been thoroughly analyzed in this paper. Based on the result, the credit risk and its causation have been deeply analyzed in the course of building the system, whose index system is consequently established.Correct evaluation of the credit risk is the key to the building of early-warning system. The risk-evaluating model based on Hopfield neural network can evaluate the risk by using the association function of network through the design of the weight to set the given risk models in the network. The difficulty to make early-warning line of the existing models has been overcome by the model. At the same time, the invalidation of the fuzzy synthetic judgment has been avoided by it.The control of the credit risk is the objective of the early-warning system. It falls into three categories: the liquidity risk control, investment risk control and loan risk control. As for the control of liquidity risk, the wavelet network method has been introduced, which can be used to control liquidity risk when precise forecasting of the time-series of economic data has been made. Investment risk can be controlled by both systematic and non-systematic risk pre-control. The systematic risk control may be realized mainly by financial futures and financial options, while the non-systematic risk by investment decentralization. The loan risk control can be realized by fuzzy synthetic judgment and the best controlling method can be derived from the optimizing model while the capital is limited. Finally, credit risk-controlling system has been studied using the radial basis function (RBF) neural networks. Combining the quantitative and qualitative methods, the system has simple algorithm and strong robustness. Simulation testifies to the feasibility of the controlling methods in above.The study indicates that the early-warning system consists of three subsystems, i.e. that of identification, evaluation and control. Based on the evaluating index system, with mathematic tools, studying methods on credit risk early-warning system are designed considering that the artificial neural networks possess nonlinear function, adaptive learning ability, parallel and distributed processing, strong robustness and fault tolerance. Realization course of those methods and the practicability of the early-warning system are also validated theoretically and experimentally in this paper.