The invention provides a cerebral apoplexy type prediction method and device; the method comprises the following steps: obtaining brain sample images of a plurality of cerebral apoplexy patients, and obtaining focus information corresponding to each brain sample image; using a machine learning method to describe the brain sample image features according to the focus information, thus forming a cerebral apoplexy focus information data model; obtaining a brain scan image, and using the cerebral apoplexy focus information data model to predict the cerebral apoplexy type of the brain scan image. The method firstly uses big data to build he cerebral apoplexy focus information model; in determination, the method needs to directly input the patient brain scan image into the model, and uses the model to calculate the brain scan image, thus finally outputting the cerebral apoplexy type; the method is fast in response speed, hard to cause condition delay, needs no artificial participation in the whole process, thus reducing doctor medical pressure; in addition, the method and device cannot have inaccurate condition determination caused by personal differences.