<![CDATA[The unpredictability of climate conditions poses significant challenges for agricultural activities, particularly in Garut Regency, where traditional knowledge often guides planting schedules. This study aims to optimize rice planting schedules by employing a Bidirectional Long Short-Term Memory (BiLSTM) model for multivariate time series forecasting of weather parameters. The research utilized meteorological data from BMKG, including average temperature, relative humidity, rainfall, and sunshine duration, which were preprocessed to ensure data quality and normalized for modeling. The BiLSTM model demonstrated superior performance in predicting key variables such as temperature and humidity, achieving low Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE). However, higher errors were observed for rainfall and sunshine duration due to the complex nature of these variables. The study successfully identified optimal planting periods by aligning weather predictions with criteria for rice cultivation, providing a comprehensive calendar to assist farmers. These findings emphasize the potential of advanced machine learning models in mitigating climate-related agricultural risks and improving productivity. Future studies may focus on integrating additional meteorological factors to enhance prediction accuracy.]]>
Copyrights © 2025
