The dispatch department of terminals has to accurately predict the length of time the ship will dock at the terminal to guide the terminal to conduct efficient resource scheduling. Also, the completion time of a ship's cargo loading or unloading process should be predicted dynamically with the progress of operation process. However, releasing prediction information that has more or less deviations from the actual values too frequently may bring obstacles to decision makings. In fact, it is possible to reduce the times of prediction information release, while not significantly losing the overall accuracy of the prediction during the whole process. In this paper, we put forward a strategy to optimize the timing of prediction information release through error analysis on the completion time predictions of container ships' operation process. The experiments on the real world data prove the superiority and practical significance of this method.