Back-propagation learning in improving the accuracy of neural network-based tourism demand forecasting

Abstract

Traditional tourism demand forecasting techniques concentrate predominantly on multivariate regression models and univariate time-series models. These single mathematical function-based forecasting techniques, although they have achieved a certain degree of success in tourism forecasting, are unable to represent the relationship of demand for tourism as accurate as a multiprocessing node-based feed-forward neural network. Previous research has demonstrated that using a feed-forward neural network can accomplish a higher forecasting accuracy than the regression and time-series techniques for a set of linearly separable tourism demand data. This research extends the applicability of neural networks in tourism demand forecasting by incorporating the back-propagation learning process into a non-linearly separable tourism demand data. Empirical results indicate that utilizing a back-propagation neural network outperforms regression models, time-series models, and feed-forward neural networks in terms of forecasting accuracy.