RESUMO
Objective To perfect the prescription knowledge discovery methods; To discover the key factors affecting the robustness of prescription therapeutic model as well as improve its recognition capability.Methods Expanded knowledge base and improved design of Chinese Medicine Prescriptions Intelligence Analytic System (CPIAS) were proposed, such as the establishment of the heuristic filtering rules of efficacy-syndrome relationship, knowledge table of efficacy-syndrome element relationship, identification of efficacy-syndrome element relationship, and syndrome element-syndrome relationship. In addition, quantitative data were calculated by CPIAS. Prescription therapeutic modeling experiments on the Chinese medicine prescriptions system were conducted based on support vector machine (CPSVM), which was also used to analyze the learning outcomes.Results Using expanded knowledge base and improved calculation results can significantly promote learning abilities of CPSVM.Conclusion Screening of efficacies, sorting of symptoms, and collection of syndrome elements are the key factors affecting the quality of prescription therapeutic model.