In this study, performance of two artificial networks was evaluated to determine which one would have more efficiency in predicting nitrate contamination of groundwater. The case study was in Babol which is recognized as one of the most fertile regions in Iran. Relevant factors including hydrogeology, soil nitrogen content, soil organic matter and soil carbon content were measured in situ as input data to predict nitrate in groundwater, then correlated by using the Pearson formula. Next, back-propagation and radial basis function neural networks were applied one-by-one. The best structure for back-propagation model was found to be 4-5-1 and Radial basis function with a spread parameter equal to 0.5 and the mean square error (MSE) of 0.50 mg/l. Results showed no significant difference between the proposed models. Both ANN models can reliably predict nitrate contamination in groundwater with acceptable accuracy. However, the radial basis model showed marginally better performance compared to back-propagation by 30 %.