Based on upstream products of FY-4A and A-Train satellites data, an estimation algorithm of cloud base height for FY-4A has been presented utilizing Random Forest model. The algorithm is evaluated in the comparison with CloudSat and CALIPSO. The results show that cloud base height for top layer cloud can be generated by using upstream products of FY-4A. Compared with CloudSat and CALIPSO, the mean absolute error is less than 1km and the relationship coefficient is bigger than 0.8. The presence of multi-layer clouds may result in underestimate of cloud base height, the error in cloud top height may also introduce uncertainties in estimation of cloud thickness and cloud base height. In addition, error of the cloud base height tend to increase as the cloud thickness increasing.