Comparison of Decomposition and Triple Exponential Smoothing Methods to Improve Rice Production Forecasting in East Java Province

This study forecasts rice production in East Java using Triple Exponential Smoothing (Holt-Winters) and Decomposition. Data includes rice production in dry milled grain (GKG) from January 2018 until December 2023, sourced from the Central Statistics Agency (BPS) of East Java. The analysis identifies the Holt-Winters Multiplicative model as the most effective, with the lowest error values: Mean Absolute Percentage Error (MAPE) of 0.1452, Mean Absolute Deviation (MAD) of 0.1078, and Mean Squared Error (MSE) of 0.0286 during training, and MAPE of 0.1974, MAD of 0.1909, and MSE of 0.0858 during testing. The Holt-Winters Multiplicative model is recommended for future rice production predictions, providing reliable method for accurate forecasting, and aiding in future rice demand planning in East Java.