Short-term forecasting for harbor waterway currents speeds

The ocean currents speeds in the harbor waterway are directly related to the ability of the ship to in or out the harbor. Accurately predict the speeds can assist the ship to choose the right time for sailing. To solve this problem, we chose two models of linear and non-linear prediction. We had set sensors in Qinhuangdao for a long time, then using the collected data for training. Our test is using a lot of random data to train and predict with different steps and orders. The results show that both methods can use less original data to train the model, and finally achieve preferably prediction. According to the characteristics of Qinhuangdao harbor, Auto-Regressive (AR) model is more appropriate than Support Vector Regression (SVR) model.