Teleoperated vehicles over 5G face reliability issues due to reactive handovers. We propose a deep learning framework using time-series radio data to predict handover onset, target cells, and link quality. These forecasts enable proactive actions like bandwidth scaling or control rate adjustment to ensure stability.