IoT-Enabled Smart Irrigation Framework Leveraging Stacked Machine Learning Algorithms

The growing need for sustainable agriculture requires innovative solutions that optimize water usage without compromising crop yield. This paper presents an Intelligent Irrigation Framework that leverages real-time sensor data from the Smart Farming Data 2024 dataset along with a robust ensemble model to enhance irrigation efficiency. The dataset includes environmental parameters such as soil moisture, ambient temperature, humidity levels, and precipitation, recorded through IoT-based sensing infrastructure. These realtime inputs are processed through a layered ensemble of machine learning models to deliver precise and reliable irrigation recommendations. Current systems lack sufficient flexibility, require high human intervention, and often rely on single-model forecasts, resulting in inefficient water management. Conventional models, such as threshold-based automation or simple regression, lack the adaptability to respond to changing field conditions. To address this concern, the proposed stack integrates Random Forest, Support Vector Regression (SVR), and LightGBM to form a hybrid prediction engine that combines the strengths of each model. Experimental outcomes demonstrate that the proposed model achieves an accuracy of 99.2% and an F1-score of 98.1%. This highperformance approach demonstrates flexibility, accuracy, and reliability in delivering smart irrigation scheduling.