A semi-distributed daily precipitation-runoff model was developed and applied to the sub-basins formed by the Barrios, Puente Paltashaco, Teodulo Peña, Malvados and Obrajillo hydrometric stations. For the development, use was made of neural networks of the Pre-fed and NARX (nonlinear autoregressive network with exogenous inputs) type, where it has been observed that the NARX network, specifically its parallel series version, has obtained the best results in the performance indicators, these being the NashSutcliffe Efficiency and the Peak Difference.