Study on Classification Model of Traditional Chinese Medicine Syndrome Types of Stroke Patients in Convalescent Stage Based on Support Vector Machine

Abstract

Objective: The purpose of this study is to construct a classification diagnosis model for different syndrome types in stroke recovery period, to provide clinical decision support for syndrome diagnosis of stroke recovery period, and to further improve clinical diagnosis efficiency and diagnostic accuracy. Method: We collected a lot of representative medical samples from medical institutions and divided them into training dataset, validating dataset and testing dataset in a 7:2:1 ratio. According to the standards of the National Administration of Traditional Chinese Medicine, we use the clinical risk characterization of common Traditional Chinese Medicine in the stroke recovery period as the input variable, and the Traditional Chinese Medicine syndrome type as the output variable. Further, we construct three diagnostic models based on Support Vector Machine for the three types of syndromes: Qi deficiency and blood stasis syndrome, phlegm-stasis in channels, and wind formation from yin deficiency syndrome. To evaluate our solution, we propose to use the TS score as a measure of our models. Results: We achieved a high classification accuracy rate of stroke syndrome: 99.54% for qi deficiency and blood stasis syndrome, 92.06% for phlegm-stasis in channels, and 85.24% for wind formation from yin deficiency syndrome. Conclusion:This diagnostic model of Support Vector Machine has a high accuracy rate. Therefore, our solution has great potential in helping clinical decision making for stroke syndrome.