Abstract: The research on dynamic evolution of tourist destination has been confined to the path of Bulter’s destination lifecycle model so that other research perspectives including fluctuation model have been neglected. Taking Huangshan Scenic Areas as a case study, this paper analyzes fluctuation characteristics of tourist arrivals by Empirical Mode Decomposition (EMD), and employs a combined forecasting model to predict tourist arrivals based on EMD and LS-SVM (Least Squares Support Vector Machines). The results show that the fluctuationof tourist arrials in Huangshan Scenic Areas present such patterns as continuously increasing trend, seasonal fluctuation, tourism cycles and economic cycles, and the combined forecasting model can predict tourist arrivals more rapidly and more accurately. All in all, EMD from fluctuation perspective can disclose dynamic evolution more directly, deeply and accurately, and combined with LS-SVM it can accurately predict tourist arrivals, which is conductive to planning management and strategic decision of scenic areas.

Key words: combined forecasting, Empirical Mode Decomposition (EMD), fluctuation of tourist arrivals, Huangshan Scenic Areas, Least Squares Support Vector Machines (LS-SVM)