Table 4.
Performance of the model in forecasting deforestation
| Id | Study area | OA | FOM | Sen | Spe | TSS | K |
|---|---|---|---|---|---|---|---|
| 1 | Andapa | 93 (0.2) | 14 (1.4) | 41 (3.4) | 94 (0.1) | 35 (3.5) | 21 (2.2) |
| 2 | Fandriana | 85 (0.3) | 20 (1.0) | 31 (1.2) | 93 (0.2) | 24 (1.4) | 25 (1.5) |
| 3 | Ivohibe | 94 (0.3) | 23 (2.4) | 41 (2.9) | 97 (0.2) | 38 (3.0) | 34 (3.2) |
| 4 | Fort-Dauphin I | 93 (0.3) | 22 (1.6) | 32 (2.1) | 97 (0.1) | 29 (2.2) | 32 (2.2) |
| 5 | Fort-Dauphin II | 96 (0.1) | 10 (1.1) | 18 (1.9) | 98 (0.1) | 16 (2.0) | 16 (1.9) |
Six performance indices were computed: overall accuracy (OA), figure of merit (FOM), sensitivity (Sen), specificity (Spe), true skill statistic (TSS), and Cohen's Kappa (K). A cross-validation procedure in which the data set was divided into training data (70%) and test data (30%) was used to compute the indices. The table lists the mean values and standard deviation of the indices for 10 repeated cross-validations.