Climate-smart agriculture using IoT-enabled machine learning for crop recommendation

Precision agriculture has emerged as a potent strategy for enhancing agricultural yield via sophisticated decision-support systems. Nonetheless, understanding soil and environmental factors improves agricultural planning. This research established a crop recommendation system utilizing an Internet of Things (IoT) methodology, accompanied by a feature for data extraction. The necessary soil nutrient data (N, P, K) was obtained using a commercial NPK sensor connected via an Arduino UNO (RS-485). Climatic Application Programming Interfaces, also called APIs, are utilized to retrieve atmospheric variables such as temperature and humidity. The system has been designed to compute an annual average instead of aggregating brief measurements, enabling it to concentrate on climatological information. The gathered dataset is subsequently analyzed using an array of machine learning methodologies including K-Nearest Neighbors, Support Vector Machines, Random Forest and Gradient Boosting. Among all the assessed models, the most effective method was Random Forest, with an experimental accuracy of 99%, thus rendering it appropriate for crop forecasting.