Price Forecasting of West Java Rice using Multivariate Decomposition SARIMAX-Gated Recurrent Unit Combination

In countries like Indonesia, rice prices significantly influence economic and social dynamics.The prices are subject to fluctuations driven by seasonal changes, market demand, and production levels, making accurate forecasting crucial.This study proposes a novel forecasting approach called Multivariate Decomposition Combination (MDC) for forecasting rice prices in West Java.This approach deconstructs a dataset into trend, seasonal, and residual components, applying multiple forecasting models.The final forecasts integrate the best-performing models for each component, enhancing overall forecasting accuracy.This study resulted in a method combination of Seasonal Autoregressive Integrated Moving Average with Exogenous Variable (SARIMAX) excelling in seasonal prediction and Gated Recurrent Unit (GRU) proficient in handling residuals and trend prediction.The combined model performs on a multivariate non-linear dataset of West Java's rice economy, achieving a Mean Squared Error (MSE) of 276,695.7,Mean Absolute Error (MAE) of 439.0, and Root Mean Squared Error (RMSE) of 526.0, outperforming deep learning individual forecasting approaches.