基于改进的BP神经网络的糖尿病肾病中医证候非线性建模研究

目的建立基于人工神经网络的糖尿病肾病(DN)证候诊断模型。方法基于MATLAB 7.0环境,采用改进的共轭梯度(trainscg)学习算法,建立DN证候三层前向BP网络模型,并用3倍交叉法验证该模型的诊断价值。结果DN证候神经网络模型预测DN证候的平均单证特异性为81.32%,平均单证准确率为96.25%,平均诊断准确率为92.21%。结论DN证候BP神经网络模型具有很好的诊断、预测能力,人工神经网络技术是中医证候非线性建模的可行性方法。

Study on non-linear modeling of TCM syndrome of DN based on developed BP neural network model

Objective To set up diabetic nephropathy(DN) syndrome diagnostic model based on artificial neural network.Method Based on MATLAB7.0 environment,the BP neural network model of syndrome was set up using the improved conjugation gradient to study algorithms and the diagnosis performance of the BP neural network syndrome model was confirmed by triple and alternative examination.Result When using DN syndrome neural network model to predict DN syndrome,the average specificity of the single syndrome is 81.32%,the average accuracy rate of single syndrome is 96.25% and the average accuracy rate of diagnosis is 92.21%.Conclusion DN syndrome neural network model is effective on diagnosis and the technology of artificial neural network is a feasible nonlinear modeling of TCM syndrome method.