<![CDATA[Rice price fluctuations in West Java significantly impact regional inflation and national food security. The main challenge in forecasting this commodity price is the high volatility of the data, which conventional statistical models often fail to capture. This study aims to compare the accuracy of four forecasting methods: ARIMA, Linear Regression, Random Forest, and Long Short-Term Memory (LSTM), to identify the best-performing model. The data used are daily medium rice prices in West Java from January 2022 to October 2024, obtained from Kaggle. The methodology includes data pre-processing, model parameter optimization, and performance evaluation using RMSE, MAE, and MAPE metrics. The results show that non-linear models significantly outperform linear models. LSTM recorded superior performance with the lowest error rates (MAPE 0.43% and RMSE 91.95), followed by Random Forest (MAPE 0.67%). In contrast, ARIMA and Linear Regression produced errors above 10%. In conclusion, Deep Learning and Machine Learning approaches are more robust at handling volatile food price data than classical econometric methods, making them highly recommended as a basis for policymakers' early warning systems.]]>
Copyrights © 2026
