Study On Syndrome Differentiation Model Of Rheumatoid Arthritis Based On Random Forest

Rheumatoid arthritis (RA) is a kind of systemic autoimmune diseases and its main performance are aggressive arthritis. TCM treatment based on syndrome differentiation is under the guidance of holistic concept, especially in preventive treatment of disease and treatment of complications, properly to avoid the inability to early diagnosis and early treatment and neglect of complications in Western medicine treatment, etc. And, compared with western medicine, traditional Chinese medicine (TCM) has many advantages such as less side effects, less adverse reaction, as well as the centralizer thrift and the ability to regulate the body’s immune function. And TCM can improve microcirculation, anti-inflammatory and analgesic, to achieve the effect of curing, making it more suitable for patients with long-term use."Syndrome" is the core of treatment based on syndrome differentiation,and finding the objective law in the theory of "Syndrome" to build specification based on syndrome differentiation and treatment is the direction of the study of TCM syndrome. The difficulty of Syndrome study lies in:First, the diversity of TCM syndrome differentiation method, and the characteristics of the card type is not standard lead to the difficulty of standardization of syndrome; Second, the traditional Chinese medicine syndrome is a complex nonlinear system, Multi-dimensional multistage and infinite combination, and it cannot be illustrated simply using the method of reduction. Third, the information of clinicians confront syndrome of decision process and results is complex and highly integration, and has the characteristics of fuzziness. Fourth, the symptom is in the identification meaning to syndrome diagnosis, lead to the difficulty in quantitative and objective of the TCM syndrome.Objective:Diagnose syndromes in TCM clinical is a process of doctor, from extraction the four diagnostic information to identify the meaningful symptoms to the process of categorizing these symptoms. Syndrome problem is essentially the classification problem of TCM symptoms. Classification algorithms are dedicated to solve the problem of classification in the field of data mining. This study introduced random forest algorithm to the study of TCM syndrome, and tried for solving the problem of feature importance calculation and classification of syndrome types of symptoms.Methods:Aim at the challenge of syndrome information in syndrome research:nonlinearity, high dimension higher-order, fuzzification and the difficulty to measure the factors important degree, this research introduced classification algorithm in the field of data mining to study the diagnosis of syndromes in TCM, used the Methods random forest for feature selection with rheumatoid arthritis and builded syndrome classification model; To test the random forest model, the research builded model using support vector machine (SVM) method as a contrast experiment,and compared the prediction accuracy of two models.Result:1. The article, with rheumatoid arthritis (RA) as the research object, collected effective treatment of RA of traditional Chinese medicine reported in literatures, classified syndrome differentiation information and standardized each terms, eventually set up a "RA certificate-symptoms" data set.2. The article constructed discriminant model with rheumatoid arthritis by adopting the method of random forests, calculated the weight of the characteristic symptoms.3. The article established the discriminant model using support vector machine (SVM) method,and compared the accuracy of two kinds of model. The result shows that random forest has a better performance.Conclusion:1. Random forests model in the process of TCM syndrome research showed good performance. Not only it had high accuracy, but also could measure the contribution of TCM symptoms in syndrome classification problem and find out the syndrome classification of the most influential main rheumatoid arthritis symptoms.2. This research adopted support vector machine (SVM) method, the relatively mature and widely used in TCM syndrome studies, to classify the same data set modeling as a contrast experiment, and the results showed that the random forest model prediction accuracy could compared with support vector machine, and the model performed more stable. The research proved that the introducing of random forest to syndrome research has significant prospects.3. A significant advantage of random forest method is it can calculate the importance of the characteristics in the course of modeling process,which embodied in this study was that realized the importance of symptom to syndrome classification of RA, selected the type to justify symptoms of identifying the most significant features, which helped to better model interpretation and strengthen the specificity of the syndrome,also provided a new method to the problem of redundancy of syndrome data and provided a new possibility to the difficulty in the study of syndrome quantitative research.