Mehmet Emin Akkaya
North Cyprus Near East University
Ihsan Tolga Medeni
Yildirim Beyazit University

Abstract
In the tourism sector, estimation in accommodation numbers is crucial for the preparation of the enterprises to answer the demand of the customers. In this study, by looking at Turkey’s three of the provinces, based on the highest tourist overnight stays, Linear Regression, MLP and SVR analyses methods has been implemented by looking at the annual data range of domestic and foreign overnight stay data. This is a comparative study for the provinces of Antalya, Istanbul, Mugla which has the highest tourist overnight stay numbers of Turkey by time series analyses performed by Linear Regression, MLP and SVR analyses methods. Annual data range has been used to estimate the tourism demand. Compared to the domestic and foreign overnight stay number estimation studies by 3 different regression methods. Used multivariate data and different tourism destinations. In the graphs with high density of cyclical fluctuation, it was observed that SVR method gave the closest result to real values. In the graphs with less density from high of cyclical fluctuation, it was observed that MLP method gave the closest result to real values. In the graphs with low density of cyclical fluctuation, it was observed that Linear method gave the closest result to real values. It is predicted that the results obtained from this study shall be useful for tourism personnel, researches, investors, tourism executives and tourism planning institutions that applies data mining techniques for tourism demand estimation applications.
Keywords :Accommodation Estimates, Tourism Forecasting, Time Series Regression