Table 4: Mean classification accuracy of each classification method against 8 different gene datasets.

DatasetICA-based RotBoostSingle TreeRotation ForestAdaBoostBaggingSVMs

Colon96.10 ± 0.5993.80 ± 0.8295.21 ± 0.4394.97 ± 0.6394.92 ± 0.5096.13 ± 0.12
CNS95.00 ± 0.2889.92 ± 0.6192.37 ± 0.8395.09 ± 0.6493.50 ± 0.7993.34 ± 0.10
Leukemia98.77 ± 0.0396.60 ± 00.4697.97 ± 0.3898.22 ± 0.5597.47 ± 0.5195.64 ± 0.49
Breast97.88 ± 0.4588.50 ± 0.7298.60 ± 0.63o98.89 ± 0.47o92.74 ± 0.4596.84 ± 0.02
Lung99.54 ± 0.1194.36 ± 0.4297.56 ± 0.2396.30 ± 0.3997.08 ± 0.3795.56 ± 0.55
Ovarian99.40 ± 0.2699.37 ± 0.1299.77 ± 0.07o99.57 ± 0.1199.76 ± 0.08o98.66 ± 0.35
MLL99.31 ± 0.5596.03 ± 0.5997.61 ± 0.3197.63 ± 0.4597.11 ± 0.5596.80 ± 0.31
SRBCT99.59 ± 0.1693.96 ± 0.5997.44 ± 0.4198.16 ± 0.3996.46 ± 0.5897.23 ± 0.44
Win Tie Loss7/1/06/0/25/2/17/0/17/1/0

Specifies that RotBoost is significantly better, and opoints out that RotBoost is notably worse at the significance level α = 0.05.