Based BP Neural Network Commercial Bank In Real Estate Credit Risk Research

After 1998, with the housing distribution system canceling and mortgage policy Implementation, individual consumption has become the main demand of real estate. China's real estate market is in the rapid development period. Real estate that is a capital-intensive industry need financial support, meanwhile the real estate is also an important service area. The primary solution for it is to improve the management of credits, to adjust the credit structure and develop the credit quality. Thus, improving the credit risk management of Chinese commercial banks deserves a deep research on it. This thesis explores the mode recognition and the relative theories and methods, propose a BP nerve net based model real estate banking credit risk estimate, recognition and early-warning.From the deep research on the formation of commercial bank's credit and specified to Chinese commercial credit management and risk classification rules, the thesis build up three parts of a real estate credit risk estimate index system, filter of this index system, and real estate credit risk general estimate model.The paper not only studies the rules of choosing risk analysis index and related issues but also set up pre-selected index group of credit risk analysis, the index include two parts, the one is about real estate enterprise financial index and non-financial index, the two index is the real estate general index system, and make the quantification to the non-financial index.Through building the index system, we can find the indicator system is too huge and complicated, affected the whole estimate results accuracy and the overall operating speed. Then, filter the general index, the main filter method is main-factor analysis. Considering explain and pertinence, use mathematical statistics in the language of hypothesis testing, through test to identify discrimination not to provide additional information variables.Staring from the study on the characteristics of neutral network and artificial NN, the research focuses on tri-hierarchy-single–unit output BP neutral network fuzzy evaluation model on credit assurance risk and the tri-layer-unit output BP neutral network model of credit risk classification.