Meteorological data influence on missing vessel type detection using deep multi-stacked LSTM neural network

Abstract

Highly-loaded seaports have extremely complex and intensive marine vessel traffic, which generates large volumes of traffic data. Meteorological conditions and maritime vessel type influence maritime traffic and they must also be taken into account in order to train the model capable of recognizing the abnormal movement of the sea transport. Real data often misses some data values, such as type of vessel or its status. This paper reviews method of obtaining vessel traffic and meteorological data and filling missing vessel type data in Rotterdam port region. A deep multi-stacked LSTM neural network model is trained to fill the missing vessel type data. This model is trained with available vessel type data and used to predict missing values. This paper describes creation and evaluation of this model. Results of the experiment show it is expedient to use traffic data of a vessel in conjunction with meteorological data