Maritime Shipping Network Graph - a Model Derived from Vessel AIS Data Creating and evaluating a graph representation of maritime vessel traffic using AIS data

Abstract

Maritime shipping shoulders more than 90% of global trade. Data science and ML, another immensely profitable industry, currently experiences an unprecedented evolution of techniques. This project proposes novel methods that apply data science techniques to the domain of maritime shipping. The main objective of this project is to construct a graph closely modelling the global maritime shipping structure, from which analytics can be derived. The node set is constructed using a pipeline of Change Point Detection to identify preliminary waypoints and reduce data quantity, KDE is utilised for geographical density estimation and partitioning the AIS data into different density areas, and lastly, the geospatial indexing framework S2 Geometry is used for final waypoint extraction to a node set. The edge set is constructed using a transition matrix that is used together with the final node set to construct the graph representation. Simulation results on the graph representation reveal the ability to construct routes with high resemblance to real-world routes. Further testing revealed high likeness between the most influential nodes in the graph representation and influential points-of-interest in the maritime shipping structure. In turn, the maritime shipping network graph representation is a tool for analysing the maritime shipping structure.