The impact of the number of training samples and the number of test points on the accuracy of land use and land cover maps derived from satellite imagery using machine learning-based classification algorithms: A case study on Al-Ahsa Oasis in Saudi Arabia

Support Vector Machines (SVM) emerged as the superior and more stable machine learning classifier for land use and land cover mapping in Al-Ahsa Oasis, Saudi Arabia, compared to Random Forest (RF), particularly when using limited training samples. The study found that SVM maintained high accuracy with as few as five training samples per class, achieving kappa values between 0.93 and 1.00 and overall accuracy from 0.95 to 1.00. Even with fewer check points, SVM's performance remained robust, contrasting RF's greater variability and lower accuracy under similar conditions. Analysis of water surface areas further highlighted SVM's consistency, with errors consistently below 5%, while RF's errors decreased significantly only as training samples increased from five to thirty. These findings underscore SVM's reliability for precise LULC mapping in this UNESCO World Heritage Site, offering critical insights for conservation and management efforts.

Faqeih, K. y. (2024). The impact of the number of training samples and the number of test points on the accuracy of land use and land cover maps derived from satellite imagery using machine learning-based classification algorithms: A case study on Al-Ahsa Oasis in Saudi Arabia. Journal of the Faculty of Arts, 84(4). https://doi.org/10.70216/2682-485x.1671