Application of Linear and Nonlinear Seasonal Autoregressive Based Methods for Forecasting Grojogan Sewu Tourism Demand

This paper discussed the application of linear and nonlinear seasonal autoregressive based methods to the monthly tourism demand in Grojogan Sewu. The linear seasonal method, i.e. Single Seasonal Autoregressive Integrated Moving Average (SARIMA) and Two Level SARIMA (TLSARIMA), and nonlinear seasonal method, i.e. Nonlinear Autoregressive (NNAR) and Two Level NNAR (TLSNNAR) were considered as appropriate methods to handle seasonal pattern in the discussed time series. In this study, the results showed that TLSARIMA and TLSNNAR are more accurate than SARIMA and NNAR in term of Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Mean Absolute Scaled Error (MASE). Based on the further examine using Diebold Mariano test, it can be concluded that the 6 periods ahead forecast values obtained by TLSARIMA and TLSNNAR have similar performance accuracy.