When a ship experiences a loss of position reference systems, the ship's navigation system typically enters a mode known as dead reckoning to maintain an estimate of the position of the ship. Commercial systems perform this task using a state estimator that includes mathematical model knowledge. Such a model is non-trivial to derive and needs tuning if the dynamic properties of the vessel change. To this end we propose to use machine learning to estimate the horizontal velocity of the vessel without the help of position, velocity or acceleration sensors. A simulation study was conducted to show the ability to maintain position estimates during a Global Navigation Satellite System outage. Comparable performance is seen relative to the established Kalman Filter model-based approach.