ARIMA Time Series Modeling with the Addition of Intervention and Outlier Factors on Inflation Rate in Indonesia

Abstract

Extreme events in a time series model can be detected when the precise timing of the event, known as the intervention, is known. When the exact timing of an event is unknown, it is referred to as an outlier. If these factors are neglected, the model's accuracy will be affected. To overcome this situation, it is possible to add the intervention or outlier factor into the time series model. This study proposes the combination of intervention and outlier analysis in time series models, especially ARIMA. It is intended to minimize the residuals and increase the accuracy of the model so that it is suitable for forecasting. Using the data of inflation rate in Indonesia, the conflict between Russia and Ukraine was used as an intervention factor in this case. Pre-intervention data (before February 2022) is used to construct the ARIMA model (1st model). After that, the modeling process continued by adding the intervention factor to the ARIMA model. The effect caused by the intervention allows an outlier to appear, so the process is continued by adding the outlier factor, called an additive outlier, into the model before (2nd model). The MAPE for the first and second models is 7.96% and 7.57%, respectively. The finding of this research shows that the ARIMA model with intervention and outlier factors, named as the 2nd model, is the best model. This study shows that combining the intervention and outlier factors into ARIMA model can improve the accuracy. The forecasting of the inflation rate in Indonesia for one period ahead in 2023 is in the range of 2.06%.

Keywords

• arima
• inflation
• economy
• forecasting
• accuracy.