Research On Intelligent Methods Of TCM Asthma Diagnosis Based On Big Data

In recent years,with the rapid development of large-capacity storage devices,wearable technology and medical and health industry informatization,the growth rate of TCM medical record data is unprecedented.Due to the high dimension,complexity and non-linearity of TCM asthma medical record data,it is difficult to establish an accurate TCM disease prediction model by using traditional machine learning algorithms,mathematical statistics and other methods.Compared with traditional machine learning algorithms,deep learning can automatically learn complex data representation from a large number of data,and has higher accuracy and generalization ability.Therefore,this research mainly studies the intelligent methods of TCM diagnosis and treatment based on deep learning,mainly including the identification and classification of the Yin and Yang of the thenar palmprint in the main symptoms of asthma and the classification of the symptom-syndrome type of asthma.The main work are as follows:(1)There is a direct correlation between the Yin and Yang characteristics of thenar palmprint and the incidence of asthma.Therefore,this research proposes an optimized YOLO V3 model to identify the Yin and Yang thenar palmprint in the palm pictures.Firstly,the improved k-means algorithm was used to cluster the palmprint image data set to determine the optimal Anchor prior parameters of YOLO V3 and optimize the model parameters.The method of data enhancement is proposed to blur,flip and scale the image data to enlarge the data volume and further improve the performance of the network.The experimental results show that the accuracy of YOLO V3 model in the classification of Yin and Yang of the palmprint reaches 92.5%.(2)The key technology of data mining is to build standardized data sets.In order to achieve the objective description of text medical record data,this research first used Pandas,a Python data analysis package,to preprocess text medical record data and extract the 20 main symptoms that affect asthma attacks.Then Map function was used to quantify the extracted main symptoms.Finally,the data are normalized by the min-max method.(3)In order to solve the problems of difficulty for feature extraction,low accuracyand easy overfitting of traditional algorithms,this research proposes an improved deep belief net(DBN)model.The classification performance of the model was optimized through the SVM classifier at the last level of the traditional DBN model.The model first learned and extracted the characteristics of the input standardized medical record data through the DBN network,and then input the extracted feature data into the cascade SVM classifier to dialectically classify the syndrome types of asthma.The performance of the model is analyzed from the aspects of learning rate,iteration times,hidden layer number and hidden layer node number,so as to determine the optimal network structure.Through comparison and analysis with other algorithms,it is found that the improved DBN model has faster convergence speed and higher accuracy for dialectical classification,so as to assist doctors in the clinical diagnosis of TCM asthma.