1997 年 34 巻 2 号 p. 155-161
It has been known that the dynamic response of rolling motions of small fishing vessels in the sea exhibits nonlinearities, because of some characteristics caused by the shapes of their bodies. This paper considers a methodology for applying a Neural Network model to the structure of the above response, and the effectiveness of forecasting the above rolling motions using this model. Comparisons of perfomances of forecasting between the method using a Neural Network model and a predictor of an autoregressive (AR) model suggested that it might be more effective to use a Neural Network model than to use an AR model. It means that the method for forecasting using a Neural Network model might be more reasonable to forecast the above roll motions than a linear forecasting method.