Enterprise Risk Pre-Warning System Of Commercial Banks Based On Bayesian Networks

Banking is one of the most risky industries in the world, the eternal theme of development of banking is the preventing and managing risks. Because of the complexity of commercial bank's enterprise risk, it is difficult to build pre-warning system by traditional methods. However, Bayesian networks have complete mathematic support and are based on the probabilistic and statistics theory. It has a lot of advantages such as intuitionistic expression, strong ability for probabilistic reasoning and the characteristic of easy understanding to humans which make it as a hot topic in uncertain theory and applying in many domains now.This paper uses the Bayesian networks as a new way to analyze the topology frame of enterprise risk and builds a causal network with key risk drives and enterprise risk. After evaluating every node, we calculate the effect of every risk factor on enterprise risk and vividly show the effect through the signal lamp model of pre-warning system. Therefore, it can help commercial banks take action to deal with risks. The research problems as follows:①The dissertation summarizes the development process of risk management concept of commercial bank, the basic principle of Bayesian networks and the research status of risk management of commercial banks based on Bayesian networks.②Analyzing the enterprise risk of commercial banks, and studying on the risk source factors and metrics in accordance with market risk, credit risk, operational risk and liquidity risk. On this basis, dissertation construct a Bayesian networks which could fully reflects the enterprise risk of commercial by set child nodes for each parent node reasonable and scientific.③Supra-bayesian was applied to determine the prior probabilities, and an example was taken to illustrate the process. On the basis of traditional expert system, Supra-bayesian use mathematical tools, give each expert's advice in different weights, and get the prior probability of scientific value at last. Comparison with the traditional setting, this method does not require a large database support, complex operation, but could set a more credible prior probability for the Bayesian networks .So, it has certain advantages. ④The enterprise risk Pre-Warning system of commercial banks based on Bayesian networks is completed. Meanwhile, the dissertation added the signal lamp model, making the results can be more intuitive, and enhance the usefulness of early warning systems.⑤The example of Pre-Warning system explains that forward inference and backward inference can help policy-makers and risk managers to predict changes of enterprise risk in individual factors or the changes of various factors in enterprise risk. So, this Pre-Warning system could support their decision-making and taking further measures to mitigate risks.