In this letter, we propose a data fusion-based predictive beamforming scheme for unmanned aerial vehicle (UAV)-assisted massive multiple-input multiple-output (MIMO) communication, which involves a base station and UAV, each equipped with a massive MIMO array. We consider aircraft dynamics to track and predict the trajectory and orientation of the UAV. To improve communication and tracking performance, we propose a novel fusion of the channel and motion data of the UAV using an extended Kalman filter (EKF). Simulation results demonstrate that the proposed scheme can improve overall spectral efficiency, particularly when the number of antennas is large.