Using a neural network to forecast visitor behavior

This study compares classical time series forecasting techniques with an emerging neural network model. The issues of data integrity, accuracy, the use of appropriate error measures, and the reliability and validation of results are highlighted. The operational aspects of forecasting with a neural network are analyzed using a data set from the US National Park Service. The study shows that the Census II decomposition and the neural network technique are the most accurate models when forecasting 12 months ahead. Results indicate that the neural network model is a valid alternative to classical forecasting techniques in tourism science.