According to cloud phase discrimination theory of spaceborne polarization lidar, and use of the method of temperature threshold for spaceborne millimeter wave radar for reference, a cloud phase discrimination algorithm using CloudSat and CALIPSO Satellite data based on support vector machines (SVM) method was established. The training and testing data of samples used for establishing SVM model were mainly derived from CloudSat 2B-GEOPROF-LIDAR, CALIPSO level 2 1km cloud layer, and ECMWF auxiliary temperature data products. The discrimination result was compared with CloudSat cloud phase product retrieved by temperature threshold method, CALIPSO cloud phase product and other relevant data. The research results show that this cloud phase discrimination technique, e.g. SVM method with combined data of radar and lidar detection, has a superior accuracy and can provide better vertical distribution information of cloud phase.