Machine learning can effectively automate safety-critical geospatial data review by encoding spatial context and attribute information, achieving 5-7% accuracy gains over baseline approaches and scaling beyond manual workflows.
This paper automates the classification of changes to Electronic Navigational Charts (ENCs)—geospatial datasets critical for maritime safety—by converting complex vector data into structured formats and using gradient-boosted trees. The method achieves 90-94% accuracy on real operational data, reducing manual review burden and improving consistency in identifying safety-critical chart updates.