Developing Crop Yield Prediction with a Novel Ensemble Framework Combining Gradient Boosted Decision Trees and Attention-based Temporal Neural Networks
Keywords:
Crop Yield Prediction, Machine Learning, Gradient Boosted Decision Trees (GBDT), Attention-based Neural Networks, Temporal Analysis, Hybrid Models, Real-time Data Integration, Sustainable Agriculture, Data AugmentationAbstract
Accurate crop yield prediction is essential for sustainable agriculture and food security in the face of climate variabilityand dynamic environmental conditions. This study introduces a novel hybrid methodology combining Gradient Boosted DecisionTrees (GBDT) and Attention-based Temporal Neural Networks (ATNN) to enhance the precision, scalability, and adaptability ofcrop yield predictions. The model leverages multi-dimensional datasets, including AgERA5 climate data, phenology datasets, and soil profiles, to integrate spatial and temporal factors. Synthetic data augmentation using
References
[1] Lenar N. Safiullin, Rais F. Sabirov, Shamil M. Gazetdinov, Vladimir M. Medvedev, Ildus K. Gimatdinov, "Multimodal AI model for wheat yield
prediction using NDVI satellite data, weather time series, and soil parameters", 2026 International Conference on Artificial Intelligence for Sustainable Engineering and Innovation (AISEI), pp.1322-1327, 2026.
[2] Avneet Kaur, Raheleh Malekian, Gurjit S. Randhawa, Aitazaz A. Farooque, Ebrahim Hashemi Garmdareh, Bishnu Acharya, Rajandeep Singh, Gurpreet S. Selopal, "Advancing sustainable agriculture in Atlantic Canada through HYDRA-SE: A hybrid decision and regression stacked ensemble for yield prediction", Advanced Engineering Informatics, vol.71, pp.104433, 2026.
