Propagation modeling and prediction are key tasks of the planning process of cellular mobile telephone systems. They should make predictions as accurate as possible using the least computing time. The accuracy of models can be strengthened by using field measurements. Neural networks are capable of learning systems of measures and give results with a high processing speed. This thesis investigates the applicability of neural networks to modeling the propagation in cellular systems. Different models are proposed for macro cells and microcells, urban and rural, in all cases achieving an accuracy level similar to that of the best models of propagation and savings calculation time very remarkable