An Alternative Metric for Detecting Anomalous Ship Behavior Using a Variation of the DBSCAN Clustering Algorithm

TL;DR

This paper introduces a new anomaly detection metric using a modified DBSCAN clustering algorithm to identify unusual ship behaviors from AIS data, enhancing statistical informativeness over previous methods.

Contribution

It presents a novel anomaly metric and explores the mathematical details of a variation of the DBSCAN algorithm for ship behavior analysis.

Findings

The new metric improves anomaly detection accuracy.

The variation of DBSCAN is mathematically detailed and validated.

Enhanced statistical interpretation of anomalies.

Abstract

There is a growing need to quickly and accurately identify anomalous behavior in ships. This paper applies a variation of the Density Based Spatial Clustering Among Noise (DBSCAN) algorithm to identify such anomalous behavior given a ship's Automatic Identification System (AIS) data. This variation of the DBSCAN algorithm has been previously introduced in the literature, and in this study, we elucidate and explore the mathematical details of this algorithm and introduce an alternative anomaly metric which is more statistically informative than the one previously suggested.