Energy-Efficient Smart Agriculture Management Framework Using IoT and Ensemble Machine Learning for Intelligent Crop Health Monitoring and Resource Optimization

Abstract

Context: Smart agriculture plays an important role in addressing the challenges of food security, environmental sustainability, and sustainable farming through IoT. Farmers are facing many issues in traditional farming, such as inefficient monitoring and poor crop health detection. Traditional farming has a negative impact on crop production and sustainability. In this regard, the development of intelligent and energy-efficient agricultural monitoring systems has become crucial for contemporary precision farming applications. Objective: This study aims to monitor crop health and optimize resource use using machine learning techniques. Our proposed system intends to monitor crop conditions in real time, enhance agricultural management, and minimize unnecessary sensor utilization and communication overhead. Material/methods : In this paper, we used five machine learning classifiers to monitor crop health using sensor data. The agricultural sensor data (soil moisture, soil temperature, soil pH, humidity, air temperature, solar radiation, wind speed, NDVI index, and 5G latency) were used for crop health monitoring. The models, such as RF, XGBoost, LightGBM, and CatBoost, and soft voting were used for crop health classification into the Healthy, High_Stress, and Moderate_Stress categories. The experimental observation revealed that soft voting achieved the highest accuracy of 94.5%, macro-F1 score of 0.945, weighted F1 score of 0.945, and AUC-ROC score of 0.988 compared to the other models. The cross-validation results further validated the robustness and stability of the proposed framework. The IoT-enabled intelligent agriculture management system offers an efficient, scalable solution for real-time crop health monitoring