Abstract:
Sri Lanka entered the international tourism market in the late 1960s. Forecasting of international tourist
arrivals to Sri Lanka is immensely important and it’s a timely requirement. Hence, finding the most
appropriate forecasting technique is essential. In view of the above, objective of the study is to forecasting
international tourism flows to Sri Lanka. Annual data on arrivals obtained from annual statistical report
2013 of Sri Lanka Tourist Development Authority from 1968 to 2013. Moving Average smoothing model,
Single Exponential Smoothing (SES) model and Double Exponential Smoothing (DES) models are tested on
monthly arrivals. The best fitting model was selected by comparing Mean Absolute Percentage Errors
(MAPE). Results revels that the single Moving Average of order 2 (MA 2) model has the least MAPE which
is 19%. SES of Alpha 0.9 has the least MAPE which is 17%.But the residuals were not normally distributed.
Double Exponential of Alpha 0.8, Beta 0.9 and Alpha 0.9, Beta 0.9 has the least MAPE which is 17%. In
these model residuals were normally distributed. Double Exponential of Alpha 0.8, Beta 0.9 and Alpha 0.9,
Beta 0.9 are the suitable models for forecasting international tourist arrivals to Sri Lanka. Smoothing
techniques can be used only for forecasting one period ahead. Therefore, it is recommended to test various
trend models on forecasting tourist arrivals.
