In recent years, machine learning tools have demonstrated their potential for accurate predictions. In the field of naval hydrodynamics, Computational Fluid Dynamics (CFD) tools are widely used to calculate ship resistance during the design stage. These tools are time-costly and require powerful computational resources. The estimation of ship resistance is a key factor for the operation of a vessel, and it should be obtained in early stages of design. Nowadays, numerical tools allow the study of different hull parameters before building the full-scale ship or an experimental model, saving time and money. Potential-flow-based tools give a relatively fast prediction when waves are dominating. However, potential solvers do not give accurate predictions when friction forces are significant. Viscous solvers are more suitable for this case, although they require much higher calculation time. The hull design of a bulbous bow vessel demands testing different bulb configurations to minimize ship resistance. This can be accomplished with viscous solvers, which are highly costly in computational resources due to the big number of models to be tested. This work proposes the creation of a surrogate model using Artificial Neural Networks (ANN) for the prediction of the ship resistance for different configurations of the bulb. The ANN is trained using a large database of different bulb configurations, and their corresponding computed ship resistance obtained with a CFD solver. Results show that the surrogate model can predict the ship resistance to high degree of accuracy and significantly faster than performing the corresponding CFD simulation.