[Purpose/significance] As an important medical digital resource, medical electronic medical record con⁃tains rich clinical information and knowledge, and its in-depth knowledge mining and discovery can discover tacit med⁃ical knowledge and provide accurate prediction and decision support for medical clinics. [Method/process] This paper discusses the framework of knowledge discovery methods for electronic medical records, proposes the implementation process of electronic medical record knowledge discovery based on the weighted association rule algorithm of principal component analysis, and conducts empirical research with the inpatient electronic medical records of patients with cere⁃brovascular diseases. [Result/conclusion] It is found that the weighted association rule method proposed in this study has some optimization effect in cerebrovascular co-morbidity pattern recognition, and the association rule and its confi⁃dence level are newly changed through the weighting process. The results demonstrate the application value and pros⁃pect of this weighted association rule method in electronic medical record knowledge discovery, and also provide new ideas and methodological guidance for future research and practice.