Research On The Model Of TCM Nursing Assistant Syndrome Differentiation Based On Machine Learning

Objective Outline of China's Nursing Career Development Plan(2016-2020)states that active development of dialectical care and specialized nursing care with TCM characteristics,innovating TCM care models,and improving the level of TCM care are national requirements for the development of nursing care.The rate of clinical technology development has been increasing,and it has shown advantages and effects that cannot be ignored in the treatment of diseases.Traditional Chinese medicine emphasizes dialectical treatment,and dialectical care is also a basic rule based on the basic theory of traditional Chinese medicine.However,TCM nursing technology has been widely developed,and basic theoretical knowledge of TCM for clinical nursing staff is relatively lacking.It is also affected by personal subjective conditions,which leads to differences in the results of dialectics.Therefore,this study uses TCM nursing technology as the starting point to explore the needs and characteristics of TCM nursing technology.According to clinical applications,a dialectical method suitable for TCM nursing technology in the emergence stage may be used.Based on this,based on Zhu Wenfeng's concepts from TCM Syndrome Differentiation,we apply machine learning:Support Vector Machines,Fully Connected Networks(FCNs),and Autoencoder+FCNs to construct TCM nursing technology-assisted dialectical tools,respectively,to verify the accuracy of different models.Methods This study draws on Zhu Wenfeng's concepts form TCM Syndrome Differentiation,based on the syndrome information collection form,and collected 173,151 pieces of data on the symptoms and signs of the 297 TCM outpatients in a top-three Chinese medicine hospital from 2018 to 2019.Support Vector Machine,Fully Connected Networks,and Autoencoder+FCNs are utilized to build auxiliary dialectical tools.We use the collected data for model training and test the models through 10-fold cross-validation for comparing the accuracy of the results.Results 1.A total of 297 patients were collected in this project,including 133 males(44.8%),and 164 females(55.2%).The yongest was 19 years old,the largest was 72 years old,and the average age was 48.14 ± 10.67 years old.2.We find that in the lower dimension,the data distribution of different syndrome types is highly coincident,and the differentiation of the syndrome results is not apparent by using t-SNE to visualize the collected data.3.Among the collected data,the most frequently occurring syndromes are in the following categories:pale tongue,greasy tongue,yellow tongue coating,history of new abortion surgery,insomnia,and tongue fingerprints.4.Under the same training set and test set,Fully Connected Network achieved the best results in the three models,among which the accuracy rates of the three types of syndromes:Biao Li,Han Re,and Xu Shi are 95.86%,97.58%,and 96.55%.Besides,the modeling time of Support Vector Machine,Autoencoder+FCNs,Fully Connected Network are 0.82s,6.27s,and 23.59s,respectively,of which Support Vector Machine takes the shortest time.Conclusion Fully Connected Network has a high accuracy rate in the eight outlines of syndrome differentiation,which can assist clinical nurses in the differentiation of syndromes before the operation of TCM nursing technology,and achieve "dialectical treatment." It should further consider its operability when developing clinical applications.