Assessing the Performance of Machine Learning Models Using Satellite Dataset for Classification of Various Crop Types

This paper aims to evaluate machine learning (ML) models' performance, named Random Forest (RF), Gradient Boosting (GB), Artificial Neural Network (ANN), and Support Vector Machine (SVM) for classifying different crop types using images from the Sentinel 2 satellite. The result shows that RF demonstrated the highest values across all criteria, outperforming the other models with an overall accuracy of 0.7365 (+ 0.010). With an accuracy of 0.7183 (+0.016) overall, GB trailed closely behind. The two models demonstrated robust performance, as evidenced by their respective F1-scores of 0.7344 and 0.7161, which suggest a well-balanced compromise between recall and precision. With total accuracies of 0.5429 (+ 0.0024) and 0.5169 (+ 0.017), respectively, the ANN and SVM algorithms performed comparably worse. In comparison to ensemble techniques like RF and GB, this shows that these models might not be as effective for this specific classification problem. The findings of this study can inform decision-making in the selection of ML algorithms for various applications, such as spatial water management, specifically in agriculture and irrigation, agricultural land management, and planning to mitigate the impact of natural disasters in the agricultural sector.