Robust Framework to Combine Diverse Classifiers Assigning Distributed Confidence to Individual Classifiers at Class Level

Table 6

Comparison based on classification accuracies of proposed approach as compared to competitors in the presence of different levels of class label noise.

Dataset	Adaboost		Bagging		RSM		Proposed approach
	5% noise	10% noise	5% noise	10% noise	5% noise	10% noise	5% noise	10% noise
IRIS	98	97.8	95.6	95.6	97.6	97.6	98	97.8
DiaretDB	90.95	88.6	97.2	96.1	95.9	94.1	97.9	97.2
Satimage	75.5	72.2	91.2	90.8	90.5	90.3	93.4	92.6
Pima Indian Diabetes	75.4	74.9	77.3	76.1	76.7	74.5	78	77.7
German	71.58	69.9	72.6	71.3	69.9	68.4	73.6	72.8