<![CDATA[This paper conducts an in-depth analysis of Indonesia’s monthly inflation rate from January 2006 to December 2024, employing advanced time series techniques to uncover underlying patterns and to develop a robust predictive framework. Utilizing Python’s TimeSeriesSplit for cross-validation, we implement and compare multiple forecasting models—specifically ARIMA, SARIMA, and Exponential Smoothing—evaluating their performance across a rolling forecast horizon. The study identifies key periods of volatility linked to the 2008 global financial crisis, domestic fuel subsidy reforms, and the COVID-19 pandemic, and assesses the degree to which seasonal and trend components explain inflation behavior. The SARIMA model selection yields SARIMA(5,1,1)(1,0,1,12), with AIC = 241.576. The seasonal MA coefficient is -0.8062 (t-stat = 0,000), indicating significant seasonal persistence. The lower AIC suggests that the seasonal component improves model fit. Our findings indicate that while seasonal patterns are present, they are relatively mild, and that a SARIMA model incorporating both non-seasonal and seasonal elements yields the most accurate out-of-sample forecasts. The paper contributes a methodological template for inflation forecasting in emerging markets and offers policy-relevant insights on the predictability of Indonesian inflation under structural and shock-driven conditions.]]>
Copyrights © 2026
