engleski

Evaluating Holt's Double Exponential Smoothing and Linear Trend Forecasting of Basic Tourism Time Series in Croatia

In Croatian economy in which a significant part of revenues are due to tourism, it is interesting for policymakers to create good forecasts of tourism supply and demand components. In this paper the dynamics of three time series in Croatian tourism were considered, annual data for beds, arrivals and nights for total number of tourists in Croatia, both domestic and foreign. A very detailed statistical analysis was done for the period 1996-2006. Various Holt’ s two-parameter exponential smoothing and trend adjusted forecast models and the linear regression trend model were applied. Assuming some underlying stability in the system exists, the authors did short-medium term forecasts for the two years time horizons, up to 2008. The two exponential smoothing constants and for the Holt’ s model applied were varying, so, four models were fitted. The accuracy of forecasts was examined based on Mean Absolute Deviation, Mean Absolute Percentage Error and Mean Square Error. For all three time series the Holt’ s methods showed to be less efficient than the linear trend forecasting with all three accuracy indicators of much higher values than for the linear extrapolation.

Keywords: tourism time series; Holt's exponentially smoothed and trend adjusted forecasting model; linear trend forecasting model; MSE; MAD.

Sažetak rada tiskan je posebno u: The 4th International Conference An Enterprise Odyssey : Tourism – Governance and Enterpreneurship Proceedings ; p. 263-264. Citiran je u znanstvenim bazama: EconLit, ProQuest