One of the problems in demand management is the change in demand, which causes uncertainty in the amount of production. Therefore, companies forecast demand in order to eliminate the uncertainty. There are several ways to predict demand. Series and parallel composition structures are the most important and accurate of these methods in prediction. The purpose of this paper is to model two series and parallel structures and compare the performance of these structures with each other and with their individual models. In this study, using a combination of seasonal autoregressive integrated moving average (SARIMA) and seasonal artificial neural network (SANN) widely used models, demand forecasting has been done in two Taiwan machinery dataset and soft drinks dataset. The results show that the performance of the proposed hybrid model is more accurate than the one-component models and also in the comparison of series and parallel structures, it is observed that on average the performance of the parallel structure in the field of demand forecasting produces more accurate results.