Boxplots of the overall classification accuracies developed by randomly selecting training samples and testing samples for six times from 270 samples of each class based on ETM⁺ image.

<p>(a) Training samples = 60, testing samples = 210, number of features = 4. (b) Training samples = 60, testing samples = 210, number of features = 10. (c) Training samples = 90, testing samples = 180, number of features = 4. (d) Training samples = 90, testing samples = 180, number of features = 10. (e) Training samples = 120, testing samples = 150, number of features = 4. (f) Training samples = 120, testing samples = 150, number of features = 10.