Virtual Booth

Rice production in the Philippines is highly vulnerable to climate variability, resulting in yield losses and inefficient resource use. This study introduces GR-AI-N, a hybrid forecasting and decision-support system that combines Multiple Linear Regression (MLR), Long Short-Term Memory (LSTM) neural networks, and a Generative AI chat-bot to predict rice yields, identify optimal planting times, and provide real-time farming recommendations. Historical Geo-climatic data—temperature, precipitation, humidity, wind speed, and dew point—from 2010 to 2024 were analyzed using SPSS to construct baseline MLR models. LSTM networks were then applied to capture nonlinear and time-dependent climate–yield relationships, leading to a hybrid MLR–LSTM model to further improve prediction accuracy. Results showed strong positive correlations between rice yield and temperature, wind speed, precipitation, and dew point, while humidity exhibited a negative correlation. The MLR models achieved an average R² of 0.87 and a MAPE of 4.18 percent, while the LSTM networks exceeded an R² of 0.95 with an average MAPE of 1.25 percent. The hybrid model demonstrated even greater accuracy, aligning closely with historical yields and producing reliable optimal planting forecasts projected up to 2100. The Generative AI chat-bot also showed high accuracy based on confusion matrix metrics, while user surveys indicated excellent functionality, usability, adaptability, and acceptability. In conclusion, GR-AI-N integrates scientific rigor with practical application by providing farmers with accurate forecasts, optimized planting guidance, and AI-driven decision support for sustainable rice production. Its web-based platform further enhances accessibility and usability for agricultural communities.