Prediction of Ship Track Anomaly based on AIS data using Long Short-Term Memory (LSTM) and DBSCAN

Abstract

Anomaly is defined as an event or behaviour that deviates. Anomaly behaviour often carried out by ships sailing in Indonesian waters is to turn off AIS for a very long time. Ship identification is challenging for maritime inspectors to detect if the ship frequently turns off AIS during the voyage. Based on these problems, the researcher proposes to predict the ship's trajectory using historical AIS data. It is possible to predict ship trajectories based on AIS data to determine the trajectory of ships that turn off AIS. The results of the AIS trajectory are combined with the ship's trajectory, in general, to determine the trajectory with the ship's trajectory in general. A predicted ship trajectory that is not the same as the general route will be identified as an anomaly. Long Short Term Memory is proposed to predict ship trajectory using latitude, longitude, speed, and time parameters. LSTM modelling with four hidden layers, determining the batch size of 25, optimizing adam, epoch worth 100, and the loss function using the mean squared error. DBSCAN clustering is used to determine the trajectory with ship trajectories in general. The platform design is the ship trajectory in this study using the CRISP-DM methodology. This study evaluates the MAE and MSE values. The data used in this study is data on ships passing through ALKI 1 in July-September 2022. From this data, data on ships moving from Singapore to Tanjung Priok Jakarta are taken. The test results show that the algorithm performs well, with an MAE value of 0.0667 and an MSE of 0.0091.