PREDICTION OF VESSEL ARRIVAL TIME USING AUTO IDENTIFICATION SYSTEM DATA

摘要

The purpose of this study is to accurately predict the ship's arrival time using a data-driven methodology. In previous studies, studies have calculated the arrival time based on distance after generating a route using ship data to predict the ship's arrival time and a study that applied trajectory data and weather forecast data to machine learning methodology. These studies have a limitation in not considering the maritime situation information in predicting the arrival time. This study proposes a data-driven methodology that uses vessel trajectory data to consider different maritime situations depending on the location. In this approach, AIS data are preprocessed using the trajectory mining technique and based on this, a pathfinding algorithm is applied and arrival time is estimated. As a result of comparing this method to the actual ship data of a terminal in Busan, Korea, the average error was improved by about 20% compared to the benchmarking methodology. This information helps improve port monitoring and berth planning and is expected to increase port operational efficiency.