基于决策树及神经网络的高血压病阴阳两虚证诊断模型的研究

Study on Diagnosic Model of Syndrome of Deficiency of Both Yin and Yang in Hypertension Based on Decision Tree and Neural Network

目的:基于决策树及神经网络的方法,建立高血压病阴阳两虚证的诊断模型。方法:从古今医案及临床病例中收集高血压病病例,对所收集资料中患者的中医四诊信息进行归一化处理,并建立证候要素数据库,运用CHAID、CRT、QUEST及C5.0决策树算法和神经网络的方法提取高血压病阴阳两虚证的诊断规律,建立诊断模型。结果:采用CHAID、CRT、QUEST及C5.0决策树算法建立高血压病阴阳两虚证诊断模型,准确率分别为93.1%、91.5%、91.5%、96.03%,其中C5.0算法的诊断模型更优于其他3种;采用多层感知器(MLP,Multilayer Perceptron)、径向基函数(RBF,Radical Basis Function)建立高血压病阴阳两虚证诊断模型,前者训练样本正确百分比为95.9%,测试样本正确百分比为93.9%;后者训练样本正确百分比为99.2%,测试样本正确百分比为96.3%,径向基函数神经网络的诊断模型更优于多层感知器神经网络。结论:通过联合应用决策树及神经网络两种方法可见,腰膝酸软在高血压病阴阳两虚证的临床诊断中起决定性作用,同时结合中医四诊信息,可形成比较符合高血压病阴阳两虚证的诊断判别模型模式,为规范高血压病阴阳两虚证诊断标准提供依据。

Objective: To establish a diagnostic model of syndrome of deficiency of both Yin and Yang in hypertension based on the method of decision tree and neural network. Methods: The information of ancient and modern medical records and clinical cases of hypertension were collected. The collected four diagnostic methods information of TCM was normalized and the syndrome factor database was established and the C5.0, CRT, CHAID, QUEST decision tree methods and neural networks in SPSS 20.0 software were used for data analysis. The diagnostic rules were extracted from TCM clinical syndromes and finally the diagnostic models were established. Results: The diagnostic accuracy of CHAID, CRT, QUEST and C5.0 were 93.1%,91.5%,91.5% and 96.03%, respectively. The accuracy rate of C5.0 decision tree model was higher than that of the other three algorithms. Multi-layer perceptron neural network and radial basis functionneural network were used to analyze the original syndrome data. The training accuracy of the former was 95.9% and the test accuracy was 93.9%, while the training accuracy of the latter was 99.2% and the test accuracy was 96.3%. The diagnostic model of radial basis function neural network was better than that of the multilayer perceptron neural network. The results were consistent with the statistical requirements. The clinical diagnosis of Yin-Yang deficiency of hypertension was better guiding. Conclusion: Combination of decision tree and neural network data mining method shows weak waist plays a decisive role in the clinical diagnosis of Yin and Yang deficiency syndrome of hypertension, combining TCM four methods diagnostic information to form a combinatorial discriminant pattern that conforms Yin and Yang deficiency syndrome of hypertension to provide evidence for syndrome standardization study.