Hybrid metaheuristic and machine learning approaches for portfolio optimization in sustainable crop production and agricultural decision support

The increasing complexity of agricultural systems, driven by climate variability, resource constraints, and market uncertainty, necessitates robust computational frameworks for crop production and decision support. This study presents a structured review of metaheuristic optimization and machine learning approaches for portfolio-based crop production, with emphasis on their integration for adaptive and data-driven agricultural planning. The review synthesizes peer-reviewed literature retrieved from Scopus and Web of Science, covering developments in supervised learning, unsupervised learning, and hybrid ML–metaheuristic frameworks. The analysis shows that ensemble and tree-based models, including Random Forests and gradient boosting methods, consistently demonstrate strong predictive performance on structured agricultural datasets, whereas deep learning architectures, such as convolutional and recurrent neural networks, are more effective at capturing spatial and temporal variability in climate-driven and remote-sensing data. Unsupervised techniques contribute to dimensionality reduction and feature extraction, improving model efficiency in high-dimensional datasets. Furthermore, hybrid frameworks that integrate metaheuristic algorithms, such as genetic algorithms and particle swarm optimization, enhance model performance through optimized feature selection, parameter tuning, and multi-objective optimization. The review also identifies critical limitations, including limited model generalization across agroecological regions, a lack of standardized evaluation protocols, computational complexity, and challenges related to interpretability and real-world deployment. This study contributes by providing a comparative synthesis of methodological approaches, clarifying their application domains, and identifying cross-cutting challenges. It further outlines research priorities focused on developing scalable, interpretable, and context-aware decision-support systems that integrate multi-source data and support sustainable agricultural management.