During the past decades, the seasonal ARIMA model (SARIMA) is one of the widely used linear models in tourism forecasting. More recently, the artificial neural networks (ANNs) have been used as an alternative to the traditional linear approaches. ARIMA and ANN models are often compared with different conclusions in forecasting performance. Tourism time series often consists of complex linear and non-linear patterns and difficult to forecast. In this research, a novel hybrid approach combining both the SARIMA and ANNs models is proposed to forecast the tourism demand in Taiwan. The hybrid model would be able to strength the unique use of ARIMA and ANN models in linear and nonlinear modeling. Monthly time series data covering from 2000.01 to 2010.12 are used in this research. The mean absolute percentage error (MAPE) is used to compare the performance of the hybrid model against other two models (i.e., the SARIMA model and the ANNs model). Results show that the hybrid forecast outperforms among the three models.