Analisis Perbandingan Peramalan Indeks Harga Konsumen di Indonesia dengan Metode Autoregressive Integrated Moving Average dan Bayesian Structural Time Series

Abstrak

The Consumer Price Index (CPI) is an important macroeconomic indicator that reflects inflation and price stability in a country. This study aims to compare the forecasting accuracy of the Autoregressive Integrated Moving Average (ARIMA) and Bayesian Structural Time Series (BSTS) methods in predicting the CPI in Indonesia, as well as to provide methodological recommendations for researchers and policymakers. The data used consists of Indonesia’s monthly CPI from January 2020 to December 2024, obtained from the Central Statistics Agency (BPS), comprising a total of 60 observations. The data was divided into training data (85%) and test data (15%). The results of the study indicate that the best ARIMA model is ARIMA (0,2,1) with a MAPE value of 1.43%, projecting that the CPI is likely to decline from 105.85 to 97.34 (indicating deflation). Meanwhile, the best BSTS model is the state component semilocal linear trend with 1,000 iterations and a MAPE of 0.21%, projecting the CPI to remain relatively stable around 106. Quantitatively, BSTS demonstrates a significant advantage in accuracy compared to ARIMA, while ARIMA is superior at capturing long-term trend dynamics.  Based on these findings, it is recommended to use BSTS if the primary priority is forecast stability and the highest accuracy, whereas ARIMA is more suitable if the objective of the analysis is to capture historical trend dynamics and long-term projections.