According to the role-based cloud classification method, a SVM (Support Vector Machine) model is established to achieve cloud type classification by extracting characteristic parameters of spaceborne millimeter-wave radar sounding data.By comparing with the classification results using BP(back propagation) network model, it is found that both two models have good classification capability, but the SVM model has better identification accuracy and faster calculation speed.An example of cloud type classification based on CloudSat data shows that the results of SVM model are well consistent with products published by CloudSat DPC(Data Processing Center).