Modelling of an Efficient System for Predicting Ships’ Estimated Time of Arrival

Abstract

Ports serve as a focal point for the global economy. Around 80% of global trade logistics is carried out through ports. As such, port efficiency is an important factor that has to be maintained properly in order to ensure a maximum economic gain. In order to boost port efficiency, a smart port system integrates state-of-the-art technology with a port management system. Predicting a ship’s expected time of arrival (ETA) is a vital step in the development of a smart port system. This study aims to develop a data- driven model to estimate the ETA of incoming ships to port Klang of Malaysia based on past voyage data. An artificial neural network (ANN) based model to predict ETA has been proposed in the study which uses the remaining distance following the trajectory, the instantaneous speed and heading of the ship as input parameters. The proposed model achieves a Mean Absolute Percentage Error (MAPE) value of 36.99% with a Mean Absolute Error (MAE) value of 4603.1367 seconds and a Root Mean Square Error (RMSE) value of 14029.6972 seconds. The model’s coefficient of determination was calculated to be 78.67% indicating a satisfactory fit to the dataset. The trajectory has been predicted using a Kalman filter and this predicted trajectory has been used as the input to the neural network model in order to provide a holistic solution to the problem.