A Comparative study of sarima and prophet models for forecasting tourist arrivals to Sri Lanka

Abstract

The Tourism sector in Sri Lanka predominantly contributes to the nation’s economy, making it vital for numerous stakeholders to understand the tourist arrival patterns to make informed decisions. Thus, primary objective of this research is to model tourist arrivals to Sri Lanka from January 2009 to June 2025 and provide accurate forecasts for the upcoming year, i.e., from July 2025 to June 2026. This univariate time series modelling was done using two approaches: SARIMA and FB Prophet modelling. Forecasts were generated using the SARIMA model and Prophet model, whereas performance metrics indicated that ARIMA (1,1,0) (1,0,1)12 model predicts with better accuracy. Moreover, the visual inspections of the time series identified major disruptions beyond the seasonal fluctuations which had a long-term impact on the tourism industry.