Research Of Credit Risk Assessment Based On Information Fusion Of State-owned Commercial Bank

The moderate operation of state-owned commercial banks has very important influence on normal development of national economy. At present state-owned commercial banks face various risk factors like credit risk, operation risk, market risk and moral risk and credit risk is the most important centralized from of financial risks in China. As the core link and an important content, credit risk assessment plays vital special role for moderate operation of commercial banks. Therefore it is of important realistic meanings to positively probe into credit risk of commercial banks, especially credit risk assessment.Issuance and implementation of New Basel Capital Accord (Basel II), especially actively introduction of The Internal Rating-Based Approach offers an all-round golden rule for risk management of global banking and has far-reaching influence on risk management of state-owned commercial banks. This paper deeply analyzes the main innovative contents of New Basel Capital Accord (Basel II) and pays attention to research on influence of Basel II on credit risk management, brings forth goals, basic principle and target of credit risk assessment based on comprehensive analysis and conclusion on requirements of Basel II framework on credit risk, relatively systematically analyzes and concludes key factors of The Internal Rating-Based Approach of credit risk and put forward 10-class rating approach credit risk assessment; the paper also give prominence to the rationale of information fusion, introduces widely used theories and methods of information fusion and combines theory and method of whole information fusion process with credit risk assessment process so as to offer theoretic base for the research on credit risk assessment and establishment of model.According to connotation of credit risk and goal of credit risk assessment, the paper brings forth basic principle of establishment of credit risk indicator system, and points out that establishment of credit risk indicator system is based on credit risk assessment. Based on it, the paper scientifically analyzes the financial factor indicators of loan enterprise and makes analysis and statement on the main factors of non-financial indicators and their influences on credit risk assessment; on the basis on construction of credit risk assessment indicator system, through relevant processing of sample data, the paper distills the characteristic elements of financial indicators and non-financial indicators, determines credit risk assessment key index , and then takes the system as input factor of credit risk assessment model. Furthermore, the paper also makes comprehensive correlation analysis on non-financial indicators, proves that non-financial indicators have obvious positive influence and effect on financial indicators and credit risk assessment through empirical analysis and indicates the necessity of integrating non-financial indicator in credit risk assessment process. The establishment of credit risk assessment indicator system offers theoretic base for the scientific assessment on credit risk.Based on analysis on theoretic method of Fuzzy Neural Network (FNN) and Support Vector Machine (SVM), the paper puts forward the design methods and realization steps of SVM-FNN, and points out that SVM-FNN method can give attention to both training error and model complexity so that SVM can obtain better generalization ability. Meanwhile, the introduction of SVM in design process of FNN can participate in parameter learning of FNN; based on deep analysis on SVM integration and information fusion theories, the paper brings forth SVM integration method based on information fusion evidence theory. The paper takes data of multiple support vector classifier integration as output of model, makes final decision-making and judgment with D-S evidence theory and points out that information fusion theory can enhance accuracy of SVM integration.Based on establishment of credit risk assessment model, the paper makes empirical analysis on seven kinds of assessment models like SVM -FNN and SVM Integration based on information fusion with sample data, points out that SVM integration model based on information fusion has relative high prediction precision; the paper also validates Boosting +D-S model of SVM Integration based on information fusion with ROC curve method and the result shows this model has high accuracy of predication, better reliability and strong assessment ability; according to standard and requirements of Advanced Internal Rating Based Approach in Basel II, the paper brings forth credit risk assessment process and sets forth some adjustment thoughts in assessment process.