<![CDATA[Food security is a strategic issue that affects economic stability and community welfare, especially in ensuring the availability of rice as a staple food in East Java. Uncertainty in food planning can cause an imbalance between rice production and consumption. Consequently, a precise forecast technique is necessary to aid decision-making. The objective of this research is to forecast or predict rice needs using the Gated Recurrent Unit (GRU) model to support more effective food management. The research methods include Min-Max Scaling normalization, and data division into 80% training and 20% testing. The GRU model has two main layers with 64 and 32 neuron units, The system was trained for 100 epochs with a batch size of 32 using the Adam optimizer and the MSE loss function. The evaluation results show high performance with MAE 0.0103, MSE 0.0001, RMSE 0.0116, and R² 0.9935, indicating low error and good generalization. The Training and Validation Loss graph shows a stable learning model without overfitting. This model can be a reliable prediction tool in food planning. Implementation of the model can help the government maintain the balance of rice supply and optimize agricultural policies.]]>
Copyrights © 2025
