<![CDATA[This study develops a statistically grounded baseline for provincial food-inflation forecasting in South Sumatra. Using monthly CPI-food data from the provincial statistics office (Jan 2024 to Jul 2025), an Auto ARIMA workflow is implemented that includes data cleaning, temporal structuring, model fitting, and holdout evaluation. On the test set (n = 4), the model attains MAE = 0.4900 and RMSE = 0.7212, indicating low average deviation and competitive short-horizon accuracy. The fitted specification produces six-month-ahead forecasts with 95% confidence intervals and offers early warning signals to support price stabilization, including the scheduling of market operations, stock management, and risk-band communication. The paper provides a reproducible forecasting pipeline tailored to subnational inflation surveillance and empirical evidence on the viability of Auto ARIMA with short samples. Limitations and extensions are discussed, including adding exogenous drivers via SARIMAX and ensemble learning.]]>
Copyrights © 2026
