对商业银行信用风险监测评价的新思考

The Consideration on the Supervision and Evaluation of Credit Risk of Commercial Bank

为监测评价商业银行信用风险,在对其成因进行深入阐述后,借助聚类分析选取32个单项指标构成评价指标体系。由此建立基于BP算法的三层前向神经网络,通过网络训练,利用网络的自适应、自学习能力,自动获取合适的网络权值与阈值,并关用附加动量法加快网络的收敛速度,基于此对商业银行所面临的信用风险进行评价。仿真试验及实证研究表明了该方法的适用性与可行性。

In order to supervising the credit risk of commercial bank, the index system comprising by 32 indexes has been established by clustering based on the analysis of the credit risk's cause. Then, the three layers forward Artificial Neural Network has been found. The weights and the threshold can be gained by the adaptive and study capability through training. At the same time, the momentum has been used in order to advancing the astringency. The credit risk degree can be obtained by the network simulation. The applicability and feasibility of the way have been indicated by the simulation and the empirical study.