Demand Forecasting and Management for Container Volume in Northeast Marmara

Abstract

Demand forecasting plays a critical role in sustaining port operations, ensuring traffic order, preventing security risks, and enabling the efficient allocation of resources in ports, which constitute the backbone of maritime transportation within the global logistics system. This study aims to forecast future container demand in the Northeastern Marmara Region by examining the effects of key macroeconomic indicators, namely gross domestic product (GDP), population, and foreign trade balance, and to provide strategic insights for port capacity planning and sectoral development. To achieve this objective, panel data regression analysis is employed to identify the relationships between container demand and the selected variables. At the same time, time-series forecasting techniques are applied to predict future cargo and container volumes. In this context, ARIMA, Exponential Smoothing (ETS), TBATS, Extreme Learning Machine (ELM), and Multilayer Perceptron (MLP) neural network models are comparatively analyzed using the R statistical software. The empirical findings reveal that GDP has a statistically significant and positive effect on both total cargo and container volumes. In contrast, population and the foreign trade balance have a negative or statistically insignificant effect. Forecasting results indicate a substantial increase in container handling volumes at the Northeastern Marmara ports over the next five-year period, with the TBATS model providing the most accurate predictions among the methods applied. The study contributes to the literature by integrating panel data analysis with multiple forecasting techniques. It offers practical implications for port authorities and policymakers in developing capacity expansion strategies and improving demand-oriented port management.