Skip to main content

Table 2 Micro-averaged number of known classes and error. Means have been computed over all iterations from all cross-validation folds for every combination of dataset, classifier and query selection criteria

From: D-Confidence: an active learning strategy to reduce label disclosure complexity in the presence of imbalanced class distributions

Dataset Classifier ff.kc c.kc dc.kc ff.e c.e dc.e
Iris SVM 2.8 3.0 3.0 0.257 0.134 0.082
Iris NNET 2.8 2.7 3.0 0.14 0.164 0.05
Iris RPART 2.8 3.0 3.0 0.342 0.187 0.184
Cleveland SVM 4.9 4.8 4.9 0.451 0.473 0.45
Cleveland NNET 4.9 4.9 4.9 0.464 0.465 0.447
Cleveland RPART 4.9 4.9 4.9 0.479 0.496 0.485
Poker SVM 8.7 7.2 8.8 0.484 0.466 0.484
Poker NNET 8.7 7.8 8.8 0.526 0.49 0.49
Poker RPART 8.7 7.5 8.6 0.524 0.495 0.517
Satlog SVM 5.6 5.8 6.0 0.349 0.186 0.162
Satlog NNET 5.6 5.9 5.9 0.729 0.726 0.739
Satlog RPART 5.6 5.9 6.0 0.430 0.261 0.28
Vowels SVM 9.8 10.4 10.5 0.546 0.341 0.322
Vowels NNET 9.8 10.7 10.6 0.661 0.601 0.623
Vowels RPART 9.8 10.7 10.5 0.645 0.617 0.632