Comparison of Representative Method for Time Series Prediction

The multi-level recursive method is a new statistical prediction theory of dynamic systems. The multi-level recursive method is used to predict ship´s rolling movements with time series prediction for the first time, which is a combined multi-level recursive method characteristic. In the same system, two different statistical prediction theories are compared through simulations. The case result shows that AR method has a certain deficit because it predicts a time-varying parameter system with a preset parameter model, the multi-level recursive time series model is effective in analyzing the time-dependent characteristics of parameters, and has considerable prediction precision. The prediction error of AR method is 4.5% and the prediction error of multi-level recursive method is 2.3%. The multi-level recursive method can also be used in time series prediction of ship´s pitching, yawing and so on