Design And Implementation Of A High-Accuracy Machine Learning Model For Multi-Crop Recommendation In Smart Farming

Abstract

Precision agriculture has proved to be a successful strategy in enhancing agricultural productivity by making decisions based on data. This study aims to develop a Machine Learning based Crop Recommendation System (CRS) which suggests appropriate crops for growers according to the nutrients in the soil and environmental condition. A well-balanced dataset with 22 crop classes, 7 input features (Nitrogen (N), Phosphorus (P), Potassium (K), temperature, humidity, pH, rainfall) and 2200 samples were used to test the performance of 26 machine learning classifiers. For the models, they were evaluated based on the Accuracy, Balanced Accuracy, Precision, Recall, F1-Score and Execution Time. Experimental results showed that the Extra Trees Classifier and Random Forest Classifier performed the best in terms of Accuracy with 99.36%, with the Extra Trees Classifier being more efficient in terms of computational time, and thus would be recommended for real-time crop selection. The statistical analysis also revealed that 15 of 26 models outperform 95% accuracy showing the framework's robustness and reliability. The developed CRS helps farmers to choose the most suitable crops based on soil fertility and climatic conditions to increase productivity, optimize use of resources and sustainable farming practices. Based on this proposed framework, an accurate, scalable and computationally efficient decision support system is built that has significant potential applications for precision agriculture and smart farming.