Forecasting Indonesia's Export Values Using SETAR-GA and SETAR-Tree

This study employs two advanced nonlinear time series models, SETAR-GA and SETAR-Tree, to forecast Indonesia's export values. The SETAR-GA model integrates a genetic algorithm to optimize parameters within a self-exciting threshold autoregressive framework, while the SETAR-Tree model combines SETAR with a recursive partitioning approach to capture regime changes more flexibly. The nonlinearity in the export data was confirmed using the Terasvirta test. The performance of both models is evaluated using in-sample and out-of-sample forecasting accuracy, assessed through MAPE and RMSE. The results indicate that both models are capable of capturing nonlinear patterns in export data, with SETAR-GA showing superior forecasting performance. These findings highlight the potential of nonlinear models to improve export forecasting in emerging economies.