基于主成分和神经网络方法的违约概率应用研究-理学专业论文.docxVIP

基于主成分和神经 基于主成分和神经网络方法的违约概率应用研究 PAGE PAGE 2 Abstract Along with the trend of financial integration and the increasing of financial market volatility, commercial banks credit risk management has become the focus of attention problems by at home and abroad Governments, financial institutions and enterprises .Therefore, it is very urgent that the credit risk of commercial banks will carry out effective supervision. Based on the Shanghai and Shenzhen exchanges of the 55 Firstly we Clustery on the sample data , the samples generally divid into two situations: normal and financial failure. companies as samples by selecting relatively reasonable indicator system, the use of principal component analysis on the data and information for effective indicators of compression, which constructed the main component factor, and neural network technology (BP Neural Network), as a principal component of the corresponding input of tectonic structure of the neural network. The combination of these two methods can overcome the main component analysis of the final conclusions of subjective judgments and reduce the principal component of the neural network layer in the middle of the number of nodes, thereby enhancing the neural network performance. Finally, we use Hybrid model and pure PCA method to compare the results The empirical results show that the hybrid model identification error is small and efficient, so it is effective for the commercial banks to assess credit risks and identify a workable solution. Keywords: Credit Risk Cluster Analysis Main Principal Component Analysis Neural Networks Hybrid Model 目录 第1章 绪言?????????????????????????????????1 1．1选题意义及背景 ???????????????????????????1 1．2国外研究现状???????????????????????????? 1 1．3 国内研究现状???????????????????????????? 3 1．4各种模型的不足和本文的创新????????????????????? 4 第2章 信用风险的含义及各种模型???????????????????????5 2.1 信用风险的含义及其博弈分析?????????????????????5 2.2 信用风险度量模型综述????????????????????????6 第3章 理论模型?????