Cloud phase classification plays an important role in meteorological forecast and climate research. The image of meteorological satellite FengYun-4 (FY-4) has more channels and better resolution than FY-2. So it provides new remote sensing data for the study of the cloud phase. This study uses a brightness temperature cloud phase index to obtain cloud phase data. Thereafter, using the cloud phase data and ensemble learning algorithm, we develop a cloud phase classification model. By applying the cloud phase classification model, the predicted classification accuracy of water cloud and ice cloud are 91.69% and 76.10%, respectively.