Applying Facebook Prophet to forecast the passenger flow in a seaport

In this study, a robust model for forecasting passenger arrivals at the Vlora seaport was developed using the Facebook Prophet method. The analysis was based on monthly data for the period 2014 - 2024, with the aim of filling the gap in the Albanian literature, in which advanced forecasting approaches for seaports are lacking. Through a comparative analysis with other forecasting methods (ARIMA, Holt-Winters, Monte Carlo, and artificial neural networks), it was found that the FB Prophet model explains 96.7 % of the data variance with a MAPE of 24.95 %, suggesting high forecasting accuracy. Also, the integration of external factors (e.g., exchange rate and GDP) showed a limited impact, highlighting the importance of historical data for this approach. These results provide a basis for managerial suggestions for the Vlora seaport, which include optimizing service lines, determining the type and size of ships, deciding on pricing policies, and organizing staff, so that the port can be better integrated into the international transport network. The study has particular value and great importance, as it fills a gap in the Albanian literature, provides an advanced methodological framework and contributes to economic development and regional integration, supporting port authorities and operators in long-term infrastructure planning and in improving the quality of passenger services. Highlights First study using AI for forecasting passenger arrivals at Vlora seaport. Comparative analysis shows FBProphet outperforms ARIMA, ANN, Holt-Winters and Monte Carlo FB Prophet model forecasts Vlora seaport passenger flow with 98 % accuracy. External factors (GDP, exchange rate, COVID-19) show a modest effect. Findings support strategic planning and infrastructure investments at Vlora port.