RESUMO
OBJECTIVE: To analyze the factors influencing the therapeutic effect of "Shoulder Tri-needles therapy" in the treatment of shoulder-hand syndrome of stroke patients by using machine learning approach, so as to provide a feasibility for improving clinical efficacy. METHODS: A total of 586 stroke patients with shoulder-hand syndrome eligible for this study were involved in our machine learning experiments for classification of the influential factors. Their data including the age, gender, pulse condition, complexion, tongue quality, tongue coating, disease stage, body mass index (BMI), blood pressure, blood glucose, blood triglyceride, blood total cholesterol, smoking history, drinking history, and final outcomes were extracted from the medical record system (from Oct. of 2014 to Jan. of 2017 in the First Affiliated Hospital and Shenzhen Futian Hospital of Guangzhou University of Chinese Medicine). The single rule algorithm (1 R) was adopted to learn, followed by optimization with Repeated Incremental Pruning to Produce Error Reduction (RIPPER) algorithm, and C 5.0 decision tree algorithm. RESULTS: The accurate classification rates of 1 R, RIPPER and decision tree model were 87.37%(512/586), 95.90% (562/586), and 97.10% (569/586), respectively. The final outcomes of machine learning of this study showed that the disease stage (acute or recovery stage), complexion difference, tongue coating difference, blood pressure level, consumption of alcohol, BMI, and smoking habit were the most important factors influencing the therapeutic effect of "Shoulder Tri-needles" in the treatment of shoulder-hand syndrome of stroke patients. CONCLUSION: The disease stage, complexion and tongue identification, blood pressure level, alcohol drinking and smoking habits, and BMI are the principal factors affecting the therapeutic effect of "Shoulder Tri-needles therapy" in the treatment of shoulder-hand syndrome of stroke patients.