基于核主成分分析和BP神经网络的电子商务信用风险预警

摘要

In this paper, we established the e-commerce credit risk classification model using the kernel PCA, PSO and neural network algorithm. More specifically, we reduced the dimensionality of the e-commerce credit risk indexes using the kernel PCA, next, using the modified PSO, searched and determined the inertial weight and threshold value of the BP neural network, using the BP neural network to train the data of 13 enterprises and then tested and forecast the data of another five, and finally we classified the results of the 18 enterprises. Through the application, we verified the merit of the KPCA-MPSO-BP based model in this respect.%用核主成分分析法(KPCA)、改进的粒子群算法(MPSO)和BP神经网络构建电子商务信用风险预警模型(KPCA-MPSO-BP).首先,用核主成分分析(KPCA)对电子商务信用风险指标进行降维处理,接着用改进的粒子群算法(MPSO)对BP神经网络的惯性权重和阈值进行搜索,确定惯性权重和阈值的大小,再用BP神经网络对电子商务信用风险的13家企业的数据作为训练集,对其进行训练,用另外5家企业的数据作为测试集,对其进行测试.实验结果表明:KPCA-MPSO-BP模型预警的误差最小,说明组合模型是合理的.