Multi-Commodity Food Price Forecasting Ahead of Eid Al-Fitr in Indonesia (2026–2030): A Comparative Study of Machine Learning Algorithms Using Time-Series Data

Abstrak

Food price volatility during the Eid Al-Fitr season poses a recurring socioeconomic challenge in Indonesia, significantly affecting household purchasing power and national food security. This study develops a predictive framework for forecasting staple food price increases ahead of Eid Al-Fitr for the period 2026–2030 using three machine learning algorithms: Random Forest (RF), Long Short-Term Memory (LSTM), and Gradient Boosting Regression (GBR). The dataset comprises multi-commodity time-series records of eleven essential commodities, including rice, chicken, beef, eggs, shallots, garlic, chili peppers, cooking oil, sugar, wheat flour, and soybeans, collected from the Indonesian National Strategic Food Price Information Center (PIHPS) spanning January 2015 to December 2025. Exogenous features, including inflation rate, USD/IDR exchange rate, fuel price index, and seasonal indicators, were incorporated. A walk-forward validation scheme with a strict chronological train validation test split was employed to prevent data leakage, and a recursive multi-step forecasting strategy was adopted for generating the 2026–2030 predictions. The results demonstrate that LSTM achieved the highest predictive accuracy with a Mean Absolute Percentage Error (MAPE) of 4.32%, followed by GBR (5.87%) and RF (7.14%). The model forecasts an average price surge of 12.6–18.4% across key commodities during the 30-day pre-Eid window for 2026–2030.