Study On TCM Syndrome Classification Based On Artificial Neural Network

The classification of TCM syndromes is a method of classifying the whole functional status of different individuals during the period of illness,and is the main dialectical basis for the clinical diagnosis of diseases in TCM,and the main basis for the dialectical treatment of Chinese medicine.However,the science of TCM syndrome classification has not been sufficiently developed,even being a little question,the main reason is that the right way and method not found,get scientific data to support the appropriate public recognition of TCM syndrome classification.TCM syndrome classification is obtained by analyzing the process of clinical symptoms in traditional Chinese medicine,this process can be regarded as a nonlinear mapping process,the development of artificial intelligence provides a new possibility for TCM syndrome classification.Artificial neural network(artificial neutral network,ANN)is a new type of intelligent information processing system to construct an artificial simulation of biological neural system,can realize the nonlinear mapping function,so it can be applied to the classification of syndromes of clinical symptoms.In this paper,the "diagnostics of Chinese medicine" the dialectical basis,in which 125 kinds of syndromes and 1084 clinical symptoms as data basis,and 125 kinds of TCM syndrome based on TCM syndrome differentiation principle is divided into six categories: the eight principal syndromes,syndrome differentiation,disease differentiation and syndrome differentiation of the six meridians,weiqiyingxue syndrome and the triple energizer.The clinical symptoms of each TCM syndrome are divided into major symptoms and auxiliary symptoms,and then the syndromes are combined according to 1-7 kinds.Each combination and main symptoms constitute all the symptoms of TCM syndromes,and there are 960590 lines of data.The data were normalized and quantified,and then trained by artificial neural networks.The input ranged from 1 to 1084,representing 1084 clinical symptoms,and the output ranged from 1 to 125,representing 125 TCM syndromes.The training result is that the correct rate of training clinical symptom data is 1.1% in CPU environment,and the time is more than 10 days.The hidden layer is 1 layers,the hidden layer node is 200-300,and the activation function is Sigmoid function.In the GPU environment are training six dialectical correct rate of clinic data were more than 98%,the time was shortened to 10 minutes,than the speed in the CPU environment faster than a thousand times,which set the hidden layer of 2 layers,each layer has 100 nodes,the activation function of ReLU function.The trained network,using the C# design interface and calling in and inputting the clinical symptoms,will predict the TCM syndromes under different dialectics.