基于决策树和人工神经网络的小儿肺炎痰热闭肺证诊断模型研究

Diagnostic model of Pediatric Pneumonia with Syndrome of Phlegm-heat Obstructing the Lung based on Decision Tree and Artificial Neural Network

目的基于决策树和人工神经网络方法,建立小儿肺炎痰热闭肺证诊断模型。方法以大样本、多中心小儿肺炎痰热闭肺证病例作为数据源,运用CRT、CHAID、QUEST、C5.0决策树和多层感知器(MLP)、径向基函数(RBF)的神经网络方法建立小儿肺炎痰热闭肺证诊断模型,并结合中医理论分析模型的诊断规则。结果采用CRT、CHAID、QUEST、C5.0算法决策树建立的小儿肺炎痰热闭肺证的诊断模型,准确率为83.1%、91.0%、89.5%、93.2%。其中采用C5.0算法的决策树模型优于前3种。采用MLP、RBF算法的神经网络方法建立小儿肺炎痰热闭肺证诊断模型,准确率为92.1%和90.8%,其中采用MLP的神经网络更优于采用RBF算法的神经网络。结论运用决策树和神经网络方法,可以建立小儿肺炎痰热闭肺证诊断模型。其中痰多粘稠、痰色黄、脉滑、咳嗽、指纹紫滞为诊断中的决定要素。"痰"、"热"为痰热闭肺证的证候病机。本研究为小儿肺炎临床辨证论治提供客观依据,有力促进中医标准化进程。

Objective To establish a diagnostic model for children with pneumonia of phlegm-heat obstructing the lung syndrome based on decision tree and artificial neural network.Methods Large-sample,multi-center children with pneumonia of phlegm-heat syndrome were used as data sources to establish a diagnosis model of pediatric pneumonia of phlegm-heat obstructing the lung syndrome by using CRT,CHAID,QUEST,C5.0 decision tree and multi-layer perceptron(MLP),radial basis function(RBF)nerves and analysing model combined with the diagnostic rules of TCM theory.Results The accuracy rate of diagnostic models of pediatric pneumonia with syndrome of phlegm-heat obstructing the lung established by CRT,CHAID,QUEST and C5.0 algorithm decision trees were 83.1%,91.0%,89.5%and 93.2%.The decision tree model with C5.0 algorithm is better than other three algorithms.The neural network with MLP and RBF algorithm was used to establish a diagnosis model of pediatric pneumonia of phlegm-heat obstructing the lung syndrome.The accuracy rate was 92.1%and 90.8%.The neural network of MLP was better than the neural network of RBF algorithm.Conclusion Using the decision tree and neural network method,a diagnosis model of pediatric pneumonia with syndrome of phlegm-heat obstructing the lung can be established.Among them,sputum thick,sputum yellow,pulse slippery,cough,and purple infantile venule of index finger are the determining factors in diagnosis.’Phlegm’and’Heat’are its syndromes and pathogenesis.This study provides an objective basis for clinical syndrome differentiation and treatment of pediatric pneumonia,and is beneficial to promote the standardization process of Chinese medicine.