Table 7 ANN Settings used for modelling.

Hidden layer neurons (h)

6

–

Topology

3-6-10

fitnet(6)

Data division (in %)

70%-15%-15%

dividerand

Data division (in data points)

84-18-18

–

Network type

feed-forward back-propagation

feedforwardnet(hiddenSizes,trainFcn)

Transfer function

Hyperbolic tangent sigmoid in hidden layer; linear in output layer

tansig in hidden layer; purelin in output layer

Training function

Levenberg Marquardt algorithm

net.trainFcn = ’trainlm’;

Performance function

Mean square error

MSE

Learning function

Gradient descent with momentum weight and bias learning function

learngdm