Ship traffic flow prediction based on AIS data mining

Abstract

The ship AIS trajectory data is a record sequence of the ship's position and time. It contains a wealth of vessel navigation information, which helps to statistically analyze and predict ship traffic in a specific water area on a small time scale. At the same time, the ship AIS trajectory data is susceptible to noise and data loss in the process of collection, transmission, and analysis, resulting in a decrease in acquisition quality. In this regard, this paper determines whether the massive AIS data is abnormal and removes it, completes the noise reduction work, and uses the cubic spline interpolation to make the lost data be reconstructed. On the basis of obtaining clean data, a discriminant function is constructed to count the regularity of arrival of the ship on the observation surface, and then a time series method is used to model the ship traffic flow through the observation section at different time periods on a certain day. The simulation experiment confirms the rationality of the forecast result through comprehensive comparison with the RBF neural network model, and provides a reference for the maritime department to implement refined management.