To cope with the issue of multi-parameter and intricate calculation of classical models in predicting rain attenuation in Ka band, a method is presented to reduce the complexity by utilizing difference stationary time-series. By this method, a prediction model is established based on the difference transform of leader rain attenuation data, and parameters of stationary time-series are estimated to compute the rain attenuation at each frequency point in Ka band for achieving linear prediction. Simulations show that the prediction precision depends on prediction interval, the number of time-series and the frequency of difference. Compared with Dissanayake-Allaut-Haidara model, the proposed model has a prediction error of less than 10-3 when prediction interval is 0.1 GHz and the number of time-series is 20 with quadratic difference. Furthermore, the simulation results also indicate that the proposed method is simple and practical due to the fact that influence of polarization mode on model parameters may be ignored.