Performance Metrics and Accuracy Assessment of Machine Learning Models for Crop Recommendation
K Senthil, Gollapalli Sri Venkata Naga Mani Teja, R Srudhakerthi
Abstract
In this work, five machine-learning techniques are compared on crop recommendation. The Random Forest model achieved an accuracy of 99.09%, with both precision and recall at 99.19%. In comparison, the Decision Tree reached an accuracy of 98.33%, but demonstrated higher precision and the Support Vector Machine (SVM) also performed well but slightly less accurately than the Random Forest, with an accuracy of 97.27%. KNN showed an accuracy of 96.06% and performed well, but the ensemble approaches surpassed it. Despite being simpler and more interpretable, Logistic Regression had the worst accuracy of 95.91%. Results: These results indicate strong performance of ensemble methods on complex classification problems. Results were consistent across Decision Tree and SVM while comparable but notably weaker in KNN and Logistic Regression. The comparative analysis highlights the primacy of Random Forest on resilience and precision with an affirmation that simpler models such as Logistic Regression still have high utility in certain circumstances. Future research could focus on fine-tuning model hyperparameters, and exploring additional methodologies to improve classification performance.