This paper study in the correlation between cerebral stroke occurrence and meteorological factors, in order to cooperate with medical institutions to intervene cerebral stroke as soon as possible. It has great significance to reduce the occurrence of cerebral stroke, and can also provide ideas and directions for cerebral stroke treatment, and at the same time, it helps the medical department to allocate resources effectively and reasonably. The data source of the cerebral stroke cases is obtained from Guizhou Provincial Center for Disease Control and Prevention (Guizhou CDC), and corresponding daily meteorological data from Guizhou Provincial Meteorology Institute, which contains 10 years data analyzed, and a method for classifying the monthly cerebral stroke incidence was proposed. This paper analyzes statistic cases data firstly and utilizes support vector machine (SVM) and the random forest (RF) methods to construct cerebral stroke risk prediction model for monthly cerebral stroke. The accuracy of SVM prediction model has been greatly improved through the optimization of SVM parameters by inertia weight allocation. The SVM model performs good prediction accuracy through training the testing data set. The correlation selection of key meteorological features and cerebral stroke incidence was carried out based on RF algorithm, the optimal subset of meteorological features was selected, and the obtained optimal subset was used to train the RF model for cerebral stroke incidence prediction and the training data set was used to verify the model. The prediction model based on SVM and RF algorithm with key meteorological features have good prediction accuracy in the meteorological high risk assessment of cerebral stroke patients, which helps to identify high-risk cerebral stroke patients early and provides individual prevention and treatment of patients.