Table 9 Feature importance Sentinel2 + BandRAtio/indices/feature importance composite band sample classification after ANN Stacked Autoencoding (BR + S2_OIF + FE).
Sl. No | Classifier | Accuracy (OA) | AUC | F1 | Kappa |
|---|---|---|---|---|---|
1 | Extra trees classifier (et) | 0.943 | 0.994 | 0.943 | 0.931 |
2 | Random forest classifier (rf) | 0.936 | 0.991 | 0.936 | 0.923 |
3 | Linear discriminant analysis (lda) | 0.905 | 0.000 | 0.903 | 0.885 |
4 | Extreme gradient boosting (Xg boost) | 0.933 | 0.992 | 0.933 | 0.920 |
5 | Light gradient boosting machine (lightgbm) | 0.937 | 0.992 | 0.937 | 0.925 |
6 | Quadratic discriminant analysis (qda) | 0.922 | 0.000 | 0.921 | 0.905 |
7 | K neighbors classifier (knn) | 0.932 | 0.981 | 0.931 | 0.918 |
8 | Naïve bayes (nb) | 0.877 | 0.973 | 0.876 | 0.852 |
9 | Gradient boosting classifier (gbc) | 0.935 | 0.000 | 0.935 | 0.921 |
10 | Decision tree classifier (dt) | 0.889 | 0.933 | 0.888 | 0.867 |
11 | Ridge classifier (ridge) | 0.803 | 0.000 | 0.787 | 0.763 |
12 | Logistic regression (lr) | 0.869 | 0.000 | 0.866 | 0.842 |
13 | SVM-linear kernel (svm) | 0.712 | 0.000 | 0.6628 | 0.654 |
14 | Ada boost classifier (ada) | 0.409 | 0.000 | 0.273 | 0.279 |
15 | Dummy classifier (dummy) | 0.186 | 0.500 | 0.058 | 0.000 |
