It is necessary to develop a system that reduces the load on the marine traffic control because its work is manual and heavy. In this study, we created a ship behavior prediction model using Recurrent Neural Network (RNN) to explore the possibility of marine traffic control and ship maneuvering support by machine learning. Specifically, we predicted the position and course of a ship that would go through the bend of the Uraga Channel from 5 items (length, width, course, speed and position) and displayed on a map. It shows that the effectiveness of ship behavior prediction by machine learning has been confirmed.
