Tourism revenue forecasting: A hybrid model approach

This paper proposes a hybrid approach, combining a Box-Jenkins ARIMA methodology and a feed-forward back-propagation network. It proposes to take advantage of the forecasting strength of ARIMA and ANN models and utilizes the capturing of linear and nonlinear patterns to achieve better accuracy in forecasting Taiwan's 1961-2009 annual tourism revenue time series data. It will not only provide a comparison of prediction accuracy of annual tourism revenue between ARIMA, an artificial neural network and the proposed ARIMA-ANN hybrid model, but also reveal how the proposed ARIMA-ANN hybrid approach could outperform the ARIMA and neural network model employed for both linear and nonlinear time series data.