Forecasting Tourism Demand in Croatia: A Comparison of Different Extrapolative Methods

Abstract

The paper investigates the forecasting accuracy of different basic extrapolative methods in modelling international tourism demand in Croatia. The study compares the results of five basic time-series forecasting methods used to predict foreign tourists’ nights, namely the Naïve 2 trend, the double moving average with linear trend, the double exponential smoothing, the linear trend time and the autoregressive method. According to the diagnostic all used models show good forecasting performances, but the double moving average method performed the best forecasting performance due to the smallest value of the mean absolute percentage error.