Cost-Sensitive Radial Basis Function Neural Network Classifier for Software Defect Prediction

Table 6

Parameters of the proposed ADBBO-RBFNN algorithm.

Parameters	BBO	Parameters	RBFNN
Habitat size	40	Learning rate	0.2
Habitat modification probability	1	Momentum factor	0.1
Immigration probability bounds per gene	0, 1	Number of hidden neurons	1/2 the number of input neurons
Step size for numerical integration	1	Maximum iteration	500
Maximum immigration	1	Activation function	Gaussian activation function
Migration rate for each island	1	Number of output neurons	1 (defect or defect-free)
Mutation probability	0.005
Maximum generation	500
Habitat search dimensional rate dr	0.5