ABSTRACT
Accurate prediction of vessel traffic flow is essential for maritime navigation departments to develop waterway planning. However, machine learning algorithms have been widely applied in marine traffic flow prediction. Existing researches still exhibit limitations in various applications, such as slow convergence speed and low prediction accuracy. To solve the problems, this paper proposes an offshore traffic flow prediction model based on an Online Recurrent Extreme Learning Machine. Firstly, we preprocess the vessel AIS (Automatic Identification System) data, and input AIS data to the model for training; Secondly, we continuously and recursively update the weight matrix by using recursive least squares to enhance its overall prediction performance; lastly, we employ the trained network to predict the future short-term vessel traffic flow. The experimental results demonstrate that the method proposed in this paper exhibits a high prediction accuracy compared to the Extreme Learning Machine and Online Sequential Extreme Learning Machine methods.
