Forecast models of cardiovescular and cerebrovescular diseases in Nanjing are separately built. First we select 22 dummy variables including holidays as nonmeteorological factors according to the time distributive characters of the daily hospital visit numbers from January 2003 to July 2007. Then we choose meteorological and nonmeteorological factors with stepwise regression method so as to obtain the explanatory variables which are finally used to build forecast models based on the SVM regression method. The daily hospital visit numbers are divided into five grades and the results show that the precisions of grade prediction in cardiovascular and cerebrovascular diseases are 87.91% and 84.62% respectively. Therefore, the models perform satisfactorily and can be applied to actual predictions. 