Steming from the hypothesis that that physical tourism turnover in the medium term depends on tourism demand as well as on the development of accommodation capacity and the dynamics of tourism seasonality, three univariate prognostic models have been constructed. Based on the Box-Jenkins approach, one ARIMA model has been built up to forecast the trend of tourism demand, then one ARIMA model to forecast the trend of accommodation capacity and one ARIMA model to forecast the trend of tourism seasonality. The outputs of all three models were then subjected to linear mathematical programming with the aim of obtaining the weights for the ectrapolation of the five year dynamics of tourists' flows in Croatia.