Estimating the Potential Impact of Shocks to Tourism Demand in Recent History of Sri Lanka Using a Machine Learning Modeling Approach

This study investigates key factors influencing tourism demand in Sri Lanka over the past five decades, using a Machine Learning (ML) modeling approach. It also estimates potential losses in tourism earnings during periods of domestic and external shocks. The findings reveal that tourist arrivals were significantly affected by historical arrival patterns, macroeconomic variables such as domestic inflation and the exchange rate, and various shocks including periods of civil war, terrorist attacks, natural disasters, and financial crises. The COVID-19 pandemic was identified as the principal external shock. Among the ML models evaluated, Multiple Linear Regression (MLR) and Extreme Gradient Boosting Regression (XGBR) demonstrated the highest accuracy in predicting tourist arrivals across different historical episodes, outperforming models such as Support Vector Regression (SVR), K-Nearest Neighbor Regression (KNN), and Gradient Boosting Regression (GBR). The study successfully quantifies the loss in tourist arrivals and potential tourism earnings attributable to these shocks within a historical context.