For collecting and sorting literature on treatment of ancient and modern stroke at the entry point, frequency and charts analysis methods were applied. Modern computer technology and statistics were used to do in-depth study of stroke literature data, explore stroke differentiation of symptoms and signs and regular treatment by ancient physicians. On this basis, the unsupervised data mining technology was applied to obtain each syndrome commonly used for drug pair and combination of core. In assembling empirically drug pair and combination of core, 3 new prescriptions were obtained for each syndrome. Key words: Unsupervised data mining, stroke, medication rules, new prescription.
