Thank you for your good presentation. It is very good to see a contribution from Karadeniz Technical University. I have a question. Did you calculate one step ahead forecasts for the test set from ANNs and SARIMA? Some computer programs (like SPSS, Eviews) give only multistep forecasts for SARIMA model. It is well known that one step forecasting performance is better than multistep forecasting performance for the SARIMA model. Moreover, I want to make a comment. In your ANN model, you need to use first and second lagged variables because your data has trend besides seasonality.
Thank you for your comments and contribution. The conference was a different experience for me. I congratulate you and the organizing committee for the successful conference organization.
As for your question, I used Minitab for SARIMA. The forecast values for the test set were realized iteratively for both methods. In the ANN model, using the past seasonal periods for the data with seasonality and trend can give more successful results. Because first and second lagged variables does not yield successful prediction results, the number of past seasonal period was investigated by the proposed ANN approach.