Abstract: Aiming at defects of slow learning speed, trapped in local solution and inaccurate operating result of BP neural network in the application of credit risk assessment of small and micro enterprises, a IDGSO-BP assessment model to measure the uncertainty credit risk of small and micro business is proposed based on Glowworm Swarm Optimization algorithm(GSO)and BP neural network. This model produces a better network architecture and initial connection weights, and trains the traditional backward propagation again by training samples. The ensemble strategy is carried out for the liner programming to calculate the best weights based on the ”Minimum square error” as the optimal rule. The simulation experimental results show that the model has obvious advantages over the traditional BP neural network model, GA-BP model and GSO-BP model in terms of convergence speed and operation accuracy. Therefore, IDGSO-BP model can effectively improve the accuracy of small and micro enterprises credit risk assessment.

Key words: small and micro businesses, credit risk evaluation, discrete glowworm swarm optimization, back propagation neural network

摘要： 针对传统BP神经网络在小微企业信用风险评估实际应用中,随机初始权值和阈值导致网络学习速度慢、易陷入局部解以及运算结果误差较大等缺陷,借助群智能萤火虫(GSO)算法,提出一种基于改进离散型萤火虫(IDGSO)算法的BP神经网络集成学习算法的小微企业信用风险评估IDGSO-BP模型。该模型以BP神经网络为基本框架,在学习过程中引入离散型萤火虫算法,优化设计神经网络的网络结构与连接权值,得到一组相对合适的权值与阈值,再进行新一轮网络训练,以“均平方误差最小”为评价准则,产生网络的输出结果,以此建立小微企业信用风险评估模型。其仿真实验结果表明,该模型在收敛速度及运算精度方面较传统BP神经网络模型、遗传GA-BP模型及连续GSO-BP模型有较明显优势。因此,IDGSO-BP模型可以有效提高小微企业信用风险评估的准确性。

关键词: 小微企业, 信用风险评估, 离散型萤火虫算法, BP神经网络