Forecasting International Tourist Arrivals from Major Countries to Thailand

Abstract

Tourism industry is one of industries that are very important for Thai economy. In order to achieve effective marketing and resource planning, accurate forecasting of tourist arrivals from major countries to Thailand is necessary for Thai tourism industry. In this paper, various forecasting models are explored to forecast monthly tourist arrivals from China, Malaysia, Korea, Japan, Russia, UK and US. The proposed models include both time series models, i.e., SARIMA, Holt-Winter, and explanatory models, i.e., Multiple Regression and Feed Forward Artificial Neural Networks (FANNs). Economic factors such as income, relative price, exchange rates, and dummy variables of seasonality and news shock effect are explored to understand their effects on international tourism demand. Mean absolute percentage error (MAPE) is used for model comparison. It was found that the more advanced model like FANNs can produce high levels of forecasting accuracy (with MAPE ≤ 10% for all studied counties) and outperform simpler forms of explanatory model like Multiple Linear Regression. However, there are counties such as US and Japan that are suitable for Holt-Winter and SARIMA, respectively, due to their obvious seasonality and trends.