An explainable multimodal data fusion framework for compatibility reasoning in heterogeneous decision systems

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Decision-support systems operating on heterogeneous agricultural data require interpretable and robust methods for modeling complex interactions among crops, nutrients, and environmental conditions. Existing intercropping and crop recommendation approaches often analyze agronomic factors independently, limiting their ability to capture the multidimensional dependencies that govern crop compatibility. To address this challenge, we propose AgriCompatNet , an explainable multimodal data-fusion framework that integrates nutrient balance, nutrient complementarity, environmental resilience, and competitive conflict effects into a unified Crop Compatibility Index (CCI). The framework combines domain-informed agronomic indicators with ensemble learning to provide accurate and interpretable compatibility reasoning through SHAP-based explanations. The framework was evaluated using 2200 crop instances, generating approximately 2.4 million crop-pair combinations. Experimental results achieved an R 2 of 0.96 and an RMSE of 0.039, outperforming conventional machine-learning baselines. Ablation studies confirmed the contribution of individual compatibility indicators, while cross-validation and environmental perturbation experiments demonstrated robustness and generalization. Explainability analyses further showed that compatibility emerges from the combined influence of nutrient balance, environmental adaptability, resource complementarity, and competitive dynamics rather than any single factor. Overall, AgriCompatNet provides a transparent and reusable framework for compatibility modeling and sustainable intercropping decision support, while offering a generalizable foundation for explainable multimodal data-fusion systems.