Machine learning-integrated plant biosensors for precision agriculture: A systematic review of stress detection, pathogen monitoring and smart crop management

Plant biosensors combining biological recognition elements with electrochemical, optical and CRISPR based transduction platforms have emerged as robust field deployable tools for the detection of stress biomarkers, pathogens and phytohormones directly within agricultural settings. Sophisticated computational methods are needed to tackle the potential of these devices which generate complex high dimensional data. By following PRISMA guidelines, this systematic review critically assesses the state of the art in machine learning algorithms like deep learning, random forest and time series models. These are used to convert raw biosensor data into actionable agronomic insights that provide early stress prediction, multi analyte classification and real time crop health assessment. The review provides a thorough survey of biosensing platforms including electrochemical, fluorescent, aptamer based, antibody based and multiplexed architectures as well as evaluates their coupling with Internet of Things (IoT) networks and artificial intelligence-based decision systems to enable precision agriculture. Some of the primary challenges are the lack of labelled field datasets, the complexity of making cross-crop models work for all crops and the lack of consistent machine learning benchmarking pipelines for biosensor data, though strategies such as transfer learning, self-supervised learning and data augmentation offer promising paths forward. The review discusses about future chances to put lightweight machine learning models directly into biosensor platforms for edge computing and wearables. This work is in line with the UN SDGs, especially SDG 2, SDG 12 and SDG 13 by integrating biosensor innovation with data-driven analytics for sustainable, climate-resilient agriculture.