ABSTRACT
This work aims at facing the challenge of providing a truthful modelling of the ship speed, vital for traffic modelling and prediction. The simulations of ship operations done with a previously developed routing software provided the large set of realistic navigation data to imitate the data set collected on-board. The data combines the considered ship speed and course, and corresponding sea conditions, as well as the propulsion configuration. Subsequently, a Bayesian Network was created with a predefined structure and its parameters were learned from data. The performed test confirmed the satisfactory level of modelling accuracy.
