Table 4. Evaluation summaries for the prediction on individual biomes on the test set for the Random Forests classifier.
Overall accuracy 0.86 | Predicted biomes | Evaluation metrics | ||||||||
---|---|---|---|---|---|---|---|---|---|---|
Overall kappa 0.75 | DXS | FGS | MGS | TSMBF | TSGSS | Recall | Precision | F1 | Kappa | |
Observed biomes | Deserts and Xeric Shrublands | 22 | 0 | 1 | 0 | 1 | 0.73 | 0.92 | 0.81 | 0.76 |
Flooded Grasslands and Savannas | 2 | 0 | 0 | 0 | 0 | 0.00 | 0.00 | - | - | |
Montane Grasslands and Shrublands | 1 | 0 | 10 | 1 | 0 | 0.77 | 0.83 | 0.80 | 0.77 | |
Tropical and Subtropical Moist Broadleaf Forests | 0 | 0 | 2 | 27 | 2 | 0.93 | 0.87 | 0.90 | 0.86 | |
Tropical and Subtropical Grasslands, Savannas, and Shrublands | 5 | 0 | 0 | 1 | 36 | 0.92 | 0.86 | 0.89 | 0.83 |
Number of correct predictions run diagonally and are highlighted in bold. Recall for for each predicted vegetation type is calculated as the weighted number of correct predictions for a given known vegetation type. Precision for each predicted vegetation type is calculated as the weighted proportion of correctly classified vegetation unit to the sum of all predictions.