A Novel Approach for Knowledge Discovery from AIS Data: An Application for Transit Marine Traffic in the Sea of Marmara

Abstract

This paper addresses the discovery of hidden patterns in the data of Automatic Identification Systems by a novel clustering model using data processing and data mining methods. It reveals the transit tracks and the transit vessels on these tracks in the Sea of Marmara which has a dense marine traffic. In this study, improved Density Based Spatial Clustering of Applications with Noise and KMeans++ clustering algorithms have been used together with complex database queries. This proposed approach has been compared to other clustering algorithms such as Self-Organizing Map, Hierarchical Clustering with Single-Link and Genetic Clustering. It has been observed that these alternative algorithms could not reach high accuracy values and they could not give the expected tracks. The proposed approach has five steps and experimental results demonstrate that when this novel approach has been applied step by step, the results can match the observed data by The Republic of Turkey, Ministry of Transport, Maritime and Communications by 95%. Finally, this novel approach is suggested to maritime authorities for all the seas in the world to manage the vessel traffic which has big and complex data.

Keywords

• clustering algorithms
• genetic algorithms
• knowledge discovery
• machine learning
• radar signal processing.