The Study Of Neural Networks In The Credit Assessment Of Individual Housing Loans

These years with the rapid development of personal housing loan operation in commercial bank, the construction of personal credit system is far from meeting the demands of consumer credit's development. Facing the absence of credit, our commercial banks lack an effectual quantitative analysis to build the personal credit scoring model. It leads to credit in arrears happens at times, the personal credit risk appeared gradually. So to study the theory and practice of credit scoring model can do well in perfecting the credit risk management of the commercial banks, promoting the growth of personal consumer credit and pushing the construction of personal credit system. It has important realistic meaning and theory value.The dissertation focuses on the problem of personal housing loan credit. It starts from the basic concepts related to credit, and then summarizes the background of the evaluating method. After that, based on referring the experts and scholasitcs'research from national and overseas, and combined with the quantitative analysis of relationship between housing loan datum and credit risk, the dissertation built a personal housing credit evaluating index system, which suits for the situation of our country. Afterwards, on the basis of three thousand personal consumers'loan datum of a national commercial bank, the dissertation utilized the Radial Basis Function (RBF for short) neural network data mining method to build an initial credit evaluation model. Then, by using a new training group of the boundary datum, which is picked from the prediction result of the first model, the paper built an optimized RBF neutral network model. According to the loanee's information, economic situation and loan's index, this model can quickly analyze the risk level of the loan. After compared with the other two data mining methods, it has been proved that the optimized model takes on well predict accuracy and shorter operation time. Finally, by using the data mining software, the paper explained the realism meaning of the predict results, and proposed some reasonable suggestions for avoiding the risk of commercial credit and construction of the national credit evaluation system.