TABLE 4.

Train parameter settings for neurofuzzy logic (FormRules^® v4.03) and artificial neural networks (INForm^® v5.01) software.

FormRules® v4.03	INForm® v5.01
Minimization parameters (ASMOD)	Number of inputs: 14
Ridge Regression Factor: 1e–6	Number of hidden layers: 1
Model Selection Criteria	Number of nodes: 2–4
Structural Risk Minimization (SRM) C1 = 0.868 (except SN: 0.8); C2 = 4.8	Transfer function in hidden layer: asymmetric sigmoid
Number of Set Densities: 2	Output transfer type: linear
Set Densities: 2. 3	Back propagation type: RPROP
Adapt Nodes: TRUE	Targets
Max. Inputs Per SubModel: 4	Target interactions: 1000
Max. Nodes Per Input: 15	Target MS error: 0.0001
	Random seed: 10000
	Test data
	Smart stop: enabled
	Minimum interactions: 20
	Test error weighting: 0.1
	Auto weight
	Interaction overshoot: 200