Research Article

Stamps Detection and Classification Using Simple Features Ensemble

Table 3

Classification performance [%].

Classifier SimpleSSigFDPDH

Bayes network (K2 learning rule)91.983.387.481.7
MLP (single hidden layer)95.794.794.188.9
SVM (Sequential Minimal Optimization, polynomial kernel)95.084.691.782.1
1NN (1-Nearest Neighbor, Euclidean Distance)96.695.595.289.9
KStar 97.095.194.090.3
Bagging (Fast Decision Tree Learner)95.491.693.589.0
RandomCommitee (RandomTree Classifiers)97.795.995.492.1
RotationForest (C4.5 Decision Tree, Principal Components Analysis)97.396.195.690.4
Nearest Neighbor with Generalization96.491.893.889.5
RandomForest (10 trees)97.495.095.190.4