Tourism Demand Forecasting for the Tourism Industry: A Neural Network Approach

Abstract

Practical tourism forecasters concentrate primarily on quantitative causal-relationship and time series methods. Although these traditional quantitative forecasting techniques have attained a certain level of success in tourism research, it is generally unknown whether they are able to simulate the relationship of demand for tourism as accurately as multiprocessing node-based artificial neural networks (ANNs). This research attempts to incorporate ANNs to develop accurate forecasting techniques for international demand for travel to a particular destination. In this study the destination featured is Hong Kong and historical data of arrivals for the period of 1970 to 1999 from Japan, UK, USA, and Taiwan, and 1984 to 1999 from China. These five countries/origins have consistently produced the largest number of inbound tourists to Hong Kong. Comparing the forecasting accuracy with five commonly used tourism forecasting techniques, we found that the ANN and single exponential smoothing forecasting models outperformed other models in terms of the chosen dimensions. Apart from its direct relevance to Hong Kong, this research provides the basis of an accurate forecasting technique that can be applied to any other travel destination.