Research Article
Stamps Detection and Classification Using Simple Features Ensemble
Table 3
Classification performance [%].
| Classifier | Simple | SSig | FD | PDH |
| Bayes network (K2 learning rule) | 91.9 | 83.3 | 87.4 | 81.7 | MLP (single hidden layer) | 95.7 | 94.7 | 94.1 | 88.9 | SVM (Sequential Minimal Optimization, polynomial kernel) | 95.0 | 84.6 | 91.7 | 82.1 | 1NN (1-Nearest Neighbor, Euclidean Distance) | 96.6 | 95.5 | 95.2 | 89.9 | KStar | 97.0 | 95.1 | 94.0 | 90.3 | Bagging (Fast Decision Tree Learner) | 95.4 | 91.6 | 93.5 | 89.0 | RandomCommitee (RandomTree Classifiers) | 97.7 | 95.9 | 95.4 | 92.1 | RotationForest (C4.5 Decision Tree, Principal Components Analysis) | 97.3 | 96.1 | 95.6 | 90.4 | Nearest Neighbor with Generalization | 96.4 | 91.8 | 93.8 | 89.5 | RandomForest (10 trees) | 97.4 | 95.0 | 95.1 | 90.4 |
|
|