Author(s): Zainab Sami Yaseen,Sanaa Mohammed Naeem & Shaymaa Qasim Mohsin
Forecasting in time series is one of the important topics in statistical sciences to help departments in planning and making accurate decisions. Therefore, this study deals with modern Forecasting methods, represented by Artificial Neural Networks Models (ANN), specifically the multi-layered network, as the Back Propagation (BP) algorithm was adopted. ) Back Propagation several times for training and selecting the lowest value of error to obtain the best model for describing the data. Classical Forecasting methods such as Box-Jenkins models were also dealt with, reconciling several models and choosing the best one for each method. These three methods were applied to realistic data on monthly average temperatures in the governorate of Basra a comparison was made between the estimated models of these methods to find the most efficient method for forecasting according to statistical measures, as it was found that the neural network method gives better and more efficient results for most time series models.