Аннотация:In the present study, a 25-year dataset was used to model rainfall-groundwater level for the next 12 months using an Artificial Neural Network (ANN) in northeastern Iran. After data preparation, 2 forward process structures and a complete data set were used to predict monthly groundwater levels. In this study, the last year's data was used as validation and the remaining data was used for network training. The results showed that using the (ANN) model in fitting and predicting monthly groundwater level for 2 structures used had a good performance and was capable of detecting trends well. In both structures, continuous overestimation and underestimation, which increases the error and reduces the performance of the models, were not observed. However, when modeling (ANN) and using the forward process structure, the model performance improved and the error amount decreased compared to the case where the complete data set was used.
