Machine Learning Models for Smart Hydroponic Systems: A Systematic Review Toward an Optimal Hybrid Modeling Frameworks

Global food production is increasingly challenged by rapid population growth, climate change, declining soil fertility, and limited arable land, particularly in developing regions. These challenges have intensified interest in soilless farming systems such as hydroponics, which offer improved resource efficiency and controlled growing environments. When integrated with sensors, Internet of Things (IoT) technologies, and automation, hydroponic systems evolve into smart hydroponic systems in which machine learning (ML) enables data-driven monitoring, prediction, and autonomous control. However, existing studies are often fragmented, focusing on isolated algorithms or narrow applications, with limited attention to scalability, robustness, and hybrid learning strategies. The aim of this paper is to systematically review and critically examine machine learning techniques applied in smart hydroponic farming systems. A PRISMA 2020-guided systematic review methodology was adopted, covering peer-reviewed studies published between 2010 and 2025 and retrieved from major scientific databases. Eligible studies were screened, quality assessed, and analyzed using structured data extraction methods. The findings show that classical supervised models such as Decision Trees, Random Forests, and Support Vector Machines perform effectively in sensor-based monitoring and control tasks, achieving accuracies of up to 98%. Deep learning models, particularly Convolutional and Deep Neural Networks, consistently outperform classical approaches in image-based applications, with reported accuracies reaching 99.7%. Hybrid ML frameworks that integrate multiple models with IoT-enabled automation demonstrate enhanced adaptability and operational efficiency. This paper concludes that while machine learning substantially improves the intelligence and performance of smart hydroponic systems, the adoption of robust hybrid frameworks, comprehensive environmental monitoring, and standardized evaluation metrics is essential for scalable, sustainable, and real-world deployment.