Skip to main content

Table 2 Characteristics of the test datasets

From: Boosting methods for multi-class imbalanced data classification: an experimental review

Dataset # of Attributes Instances # of classes IR
Conventional datasets
 Wine 13 178 3 1.47
 Hayes-Roth 4 132 3 1.7
 Contraceptive 9 1473 3 1.89
 Pen-Based 16 1100 10 2.18
 Vertebral column 6 310 3 2.5
 New thyroid 5 215 3 5
 Dermatology 34 366 3 5.6
 Balance Scale 4 625 3 5.8
 Glass 9 214 7 8.44
 Heart (Cleveland) 13 303 5 12.62
 Car Evaluation 6 1728 4 18.61
 Thyroid 21 7200 3 40.15
 Yeast 8 1484 10 92.5
 Page blocks 10 5473 5 175.46
 Shuttle 9 58,000 7 4558.6
Big datasets
 FARS 29 100,968 8 4679
 KDD Cup’99 41 494,021 5 1870
 Covertype 54 581,012 7 103
 Poker 10 1,000,000 10 64,212