Container handling forecasting with classical time series analysis

In this study, the average amount of container handling in all ports in Turkey has been forecasted. The forecast of container handling volumes will enable the planning of container flow and thus support the improvement of the service quality of the enterprises. In this thesis, data on the monthly container handling amount for the period of January 2004-December 2021 at Turkish ports were used. Although the time series used in the study is a time series with both a trend and seasonality, since it is the dominant trend in the series, predictions were made with Seasonal Naive, Holt-Winters Additive, Holt-Winters Multiplicative, ETS and SARIMA methods. Moreover, in the analysis of this time series, the model with the best test set result was determined according to the root mean squared error (RMSE) and mean absolute percent error (MAPE) criteria. As a result of the most suitable method is SARIMA, among the methods tested in this study. In addition, the average container handling prediction values of all ports in Turkey for the year 2022 were determined with 95% lower and upper confidence limits.