Intelligent Actuator Control in Smart Agriculture through Machine Learning and Sensor Data Integration

Abstract

Smart agriculture leverages Internet of Things (IoT) technology to develop intelligent greenhouses capable of monitoring and responding to environmental changes in real time. This study proposes the use of machine learning to analyze real-time sensor data—such as temperature, humidity, water level, and soil nutrient levels (N, P, K)—to determine the optimal timing for activating actuators, including fans, irrigation systems, and water pumps. In the initial stage, the study utilized the "IoT Agriculture 2024" dataset from Kaggle, which consists of 37,922 records and 13 attributes describing crop and environmental conditions. This dataset was used to train a robust machine learning model based on gradient boosting to support intelligent actuator control decisions. The model demonstrated strong predictive accuracy, achieving 99.62%. In the final stage, the model was evaluated in a simulated IoT-based agricultural system using synthetic sensor data designed to mimic real-world readings of temperature, humidity, soil moisture, and nutrient concentrations. The model achieved a high validation accuracy of 99.55%, indicating its reliability and robustness within the simulated environment. These results demonstrate that the integration of machine learning with real-time sensor data is an effective strategy for automating actuator control in smart greenhouses. The proposed approach has the potential to reduce manual intervention, optimize resource utilization, and improve overall agricultural productivity. This study contributes to the advancement of adaptive, data-driven precision agriculture systems that support long-term food security.