于GA-BP模糊神经网络的商业银行信用风险评估

Study on Credit Risk Assessment of Commercial Banks Based on GA-BP Fuzzy Neural Network

研究了商业银行信用风险评估的现状,针对单独应用BP神经网络评估信用风险时存在的缺陷,提出了基于遗传算法优化模糊BP神经网络的信用风险评估新模型。通过遗传算法训练模糊BP神经网络,克服网络建模中产生的局部极小的缺点,提高了风险评估的准确性。最后,利用Matlab软件对样本数据进行训练和测试,仿真结果表明所构造的评估模型预测误差非常小。

Studying the actuality of commercial bank credit risk assessment, its deficiency was revealed because of singly using BP neural network, and a new assessment model is proposed which is based on optmizing fuzzy neural network by genetic algorithm. The genetic algorithm is used to train fuzzy BP neural network, which can overcome the shortcoming of local minimum appeared while training the network, and the veracity of the whole assessment model can be increased. At last, the sampled data are trained and tested Matlab software, and the simulation results indicate that the proposed assessment model has very small error.