Identıfıcatıon Of Shıp Route Anomalıes On Aıs Data Usıng Dbscan Algorıthm

Abstract

Maritime is frequently used for purposes such as trade, fishing and transportation. As the intensity of use increases, the safety of maritime transportation becomes an important issue. For this reason, the positions of the ships are constantly monitored by the AIS system. AIS data includes information such as the location of the ship, the type of ship, speed, and destination port. Although AIS data is aimed to be collected regularly, it may not always be recorded properly due to system errors and GPS errors. In the analyses made, firstly, erroneous data was removed and properly recorded data were obtained. An algorithm has been developed that estimates the journey times of passenger ships with the cleaned data. From the beginning of the journey, the ship is expected to travel the same route at the same time. For this reason, the journey is divided into 10-minute segments. All experiments are applied on segment based in whole thesis. DBSCAN algorithm was used for anomaly detection. The results obtained are called possible anomalies. Possible anomaly situations were examined with weather conditions and the effect of weather on travel times was examined. Weather data is obtained based on location via OpenWeatherMap. The data includes information such as wind speed, temperature, cloud rate. As a result, it has been observed that the weather conditions have an effect on the ship's journey times. It has been determined that 45% of possible anomaly points are due to wind speed. It is aimed that all experiments can be easily repeated by people who will do similar research. For this purpose, a Jupyter Notebook was prepared and shared as open source. All analysis were divided into 10 different functions. By following these steps, it was ensured that the analyses could be repeated.