Machine learning-enabled chemical ecology for integrated pest management: from volatiles to field applications

Machine learning is transforming chemical ecology by accelerating the discovery and deployment of semiochemical-based tools for precision pest management. These advances are particularly important in the face of climate change, pesticide resistance, and the growing need for sustainable agricultural intensification. This review synthesizes how machine learning can be applied across the semiochemical discovery and implementation pipeline, from chemical signal detection to field deployment and decision support for integrated pest management. We review major machine learning approaches and demonstrate how they extract biologically relevant information from high-dimensional chemical, electrophysiological, behavioral, sensor, and field datasets. These methods accelerate semiochemical discovery, prioritize candidate compounds, optimize formulations and deployment strategies, and support adaptive pest management under dynamic environmental conditions. We further examine the integration of sensor-based technologies, explainable machine learning, and optimization algorithms to improve pest detection, monitoring, and intervention. Current challenges include data scarcity, heterogeneous datasets, the lack of benchmark resources, limited field validation, and the underrepresentation of tropical agroecosystems and smallholder farming systems. We propose practical recommendations for developing reusable datasets that integrate volatile profiles, electrophysiology, insect behavior, and management outcomes to improve model robustness and reproducibility. Central to this review is a 4-stage framework comprising detection, design, deployment, and decision support that connects machine learning with chemical ecology and provides a practical roadmap for translating computational advances into targeted, efficient, and sustainable pest management.