Predicting Vessel Trajectories from Ais Data Using R
Abstract
Analysts and security experts seek automated algorithms to predict future behavior of vessels at sea based on Automated Identification System (AIS) data. This thesis seeks to accurately predict the future location of a vessel at sea based on cluster analysis of historical vessel trajectories using a random forest. Once similar trajectories have been clustered into a route, expected prediction error can be empirically estimated based on an independent validation data set not used during training, then applied to an independent test set to produce an expected prediction region with a user-defined level of expectation. Our results show that the prediction region contains the true interpolated future position at the expectation level set by the user, therefore producing a valid methodology for both estimating the future vessel location and for assessing anomalous vessel behavior.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 2017
- Accession Number
- AD1046595
Entities
People
- Brian L. Young
Organizations
- Naval Postgraduate School