MULTI-SOURCE DATA INTEGRATION AND AI-DRIVEN DECISION INTELLIGENCE FRAMEWORK FOR PADDY YIELD PREDICTION
Abstract
Background: Paddy (rice) is one of the most important staple crops worldwide, supporting the food security of more than half of the global population. However, paddy productivity is increasingly threatened by climate variability, soil degradation, water scarcity, pest outbreaks, and unsustainable agricultural practices. Objective: Accurate and timely yield prediction is therefore essential for effective agricultural planning, resource allocation, and food security management. This study proposes a comprehensive Multi-Source Data Integration and AI-Driven Decision Intelligence Framework for paddy yield prediction. Methods: The framework integrates heterogeneous datasets including satellite remote sensing imagery, meteorological observations, soil characteristics, and ground-based biophysical parameters. Google Earth Engine (GEE) is employed for large-scale geospatial data processing. The study combines machine learning and deep learning techniques such as Random Forest (RF), Extreme Gradient Boosting (XGBoost), Long Short-Term Memory (LSTM), Convolutional Neural Networks (CNN), and Attention-based multimodal fusion models. Multi-source data fusion enables the extraction of spatial, temporal, and environmental patterns associated with crop growth dynamics. Results: Literature-based evaluations indicate that integrated AI models can achieve coefficient of determination (R²) values ranging from 0.76 to 0.90 and Root Mean Square Error (RMSE) values between 298 and 724 kg ha⁻¹, significantly outperforming conventional yield estimation approaches. The CNN-LSTM-Attention hybrid model demonstrated the highest predictive accuracy (R² = 0.90, RMSE = 320 kg ha⁻¹). Multi-source integration improved prediction accuracy by 5–15% over single-source baselines. Conclusion: The proposed framework facilitates early-season yield forecasting, risk assessment, precision resource management, and climate-resilient agricultural planning. Explainable artificial intelligence (XAI) techniques incorporated into the framework enhance model transparency and support actionable decision-making. The study demonstrates the potential of AI-enabled decision intelligence systems in advancing sustainable rice production and strengthening food security in developing agricultural economies.
Keywords
Paddy yield prediction; Multi-source data fusion; Artificial intelligence; Remote sensing; Deep learning; Decision intelligence; Precision agriculture; XGBoost; LSTM; CNN; SHAP; Google Earth Engine.