Peramalan Tingkat Inflasi di Indonesia dengan Metode ARIMA-GARCH Berdasarkan Optimasi Kalman Filter

Forecasting inflation rates is a crucial aspect of economic analysis influenced by price volatility. This volatility occurs when prices fluctuate, leading to non-constant data variance and resulting in a violation of the homoscedasticity assumption (heteroscedasticity) in inflation forecasting. Violating this assumption can cause bias in model estimation. To address the heteroscedasticity issue, this study employs the Autoregressive Integrated Moving Average (ARIMA) Generalized Autoregressive Conditional Heteroscedasticity (GARCH) model. This method can model and forecast the variance of residuals that are not constant in the data. To further optimize the model, this study integrates the Kalman Filter as a technique to minimize error covariance. The data used in this study is the inflation rate data of Indonesia from January 2010 to December 2023. Based on the analysis results, the presence of heteroscedasticity in Indonesia's inflation rate data is detected in the residuals. The best model obtained is the ARIMA(0,1,1)-GARCH(1,1) model. The application of the Kalman Filter in this method improves the estimation results, as indicated by the MAPE value of the ARIMA(0,1,1)-GARCH-Kalman Filter polynomial degree 2 at 3.60%, compared to the ARIMA(0,1,1)-GARCH(1,1) model at 12.42%. The forecasted average inflation rate for Indonesia for the next six periods ranges between 2% and 3%.