The Application of Support Vector Machines to Forecast Tourist Arrivals in Barbados: An Empirical Study

This investigation presents an SVM model with genetic algorithms to forecast the tourist arrivals and experimental results indicate that the proposed neural network outperforms the other multivariate forecasting models.

Abstract

Accurate tourist demand forecasting systems are essential in tourism planning, particularly in tourism-based countries. Artificial neural networks are attracting attention to forecast tourist arrivals due to their general nonlinear mapping capabilities. Unlike most conventional neural network models, which are based on the empirical risk minimization principle, support vector machines (SVMs) apply the structural risk minimization principle to minimize an upper bound of the generalization error, rather than minimizing the training error. This investigation presents an SVM model with genetic algorithms to forecast the tourist arrivals. Genetic algorithms (GAs) are used to determine free parameters in the SVM model. Empirical results that involve tourist arrival data for Barbados reveal that the proposed model outperforms other approaches in the literature. 1. Introduction Unpurchased service products in the tourism industry, including unfilled airline/coach seats, unused hotel rooms/hire cars, unoccupied facilities and so on, cannot be stored because they are perishable (Archer, 1987). Therefore, an accurate forecast of tourism demand, usually measured as a number of tourist arrivals, is important in assisting managerial, operational and tactical decision-making (Athiyaman and Robertson, 1992). The benefits of accurate forecasting are undisputed and various studies have been undertaken to develop the most effective tourist arrival forecasting model. Conventional quantitative forecasting models fall under two categories: regression models and time-series models (Wong, 1997). Regression models attempt to determine the relation between tourism demand and other socio-economic factors, such as income, living expenditure, and exchange rates (Lim, 1997). Such socio-economic factors, however, are not useful because their coefficients are insignificant and they contribute little to the coefficient of determination of an econometric model. However, scrapping redundant variables may increase the coefficient of determination. This fact is one of the major limitations of econometric models. Hence, more attention should be paid to collecting data concerning such variables (Sheldon and Var, 1985). Time series models are often used when data are insufficient to build econometric models or when knowledge of the structure of regression models is limited. In some cases, as in short-term forecasting, time series models are likely to outperform regression models (Morley, 1993; Witt and Witt, 1992). However, time series models cannot predict tourist arrival patterns that are not evident in historical data (Law, 2000). Artificial neural networks (ANNs) have nonlinear mapping capability and so have been recently applied to forecast tourism demand. Pattie and Synder (1996) utilized back-propagation neural networks (BPNN) model with two hidden layers to forecast monthly overnight backcountry stays in US national parks. Their forecasts are more accurate than those obtained using other traditional time series models with large samples. Law and Au (1999) proposed a feed-forward neural network with six input nodes and one output node to forecast tourist arrivals in Hong Kong. In this investigation, six factors are assumed to affect tourist arrivals. The experimental results indicate that the proposed neural network outperforms the other multivariate forecasting models. Law (2000) applied a BPNN model to forecast Taiwanese tourist arrivals to Hong Kong. The proposed neural network model contains six independent factors as input nodes, and one output node. The author used a non-linear function to separate randomly data on independent variables into training data set. This model substantially increased forecasting accuracy. Burger et al. (2001) presented eight techniques for forecasting the tourist arrivals from USA to Durban in South Africa from 1992 to 1998. They reported that the neural network approach performs best when tourist arrival data are unstructured. …