The aim of the present study is the employment of the artificial neural network (ANN) model in the estimation and evaluation of geoid heights. For that reason, a number of control points with known orthometric heights in the region of North Greece were measured by GPS and used to test the presented algorithms. The derived ANN geoid heights are compared to those produced by a wellknown conventional method through a combined gravimetric geoid. In order to evaluate the computed heights, numerical tests were carried out for points distributed inside and at the borders of the study area. The obtained results show that the ANN model is a competitive approach with certain advantages.