Automatic identification system (AIS) data based ship-supply forecasting

Abstract

Purpose: The bulk cargo shipping industry is characterized by high cost pressure. Chartering vessels at low prices is important to increase the margin of transporting cargo. This paper proposes a three-step, AI-based methodology to support this by forecasting the number of available ships in a region at a certain time. Methodology: Resulting from discussions with experts, this work proposes a threestep process to forecast ship numbers. It implements, compares and evaluates different AI approaches for each step based on sample AIS data: Markov decision process, extreme gradient boosting, artificial neural network and support vector machine. Findings: Forecasting ship numbers is done in three steps: Predicting the (1) next unknown destination, (2) estimated time of arrival and (3) anchor time for each ship. The proposed prediction approach utilizes Markov decision processes for step (1) and extreme gradient boosting for step (2) and (3). Originality: The paper proposes a novel method to forecast the number of ships in a certain region. It predicts the anchor time of each ship with an MAE of 5 days and therefore gives a good estimation, i.e. the results of this method can support ship operators in their decision-making.