Predicting Asthma Outcome Using Partial Least Square Regression and Artificial Neural Networks

Table 4

Comparison between the MLP and PNN classifiers.

Classifier	Feature size	Hidden layer		Output layer		Test success (%)
		Transfer function	Neurons	Function	Neurons	Sensitivity	Specificity	Accuracy
MLP	48	Tan-sigmoid (tansig)	6	Saturating linear (satlin)	1	100	0	83.87
PNN	48	Radial basis function (RBF); (spread = 100)	112	Competitive (compet)		92.3	60	87.09
MLP	9	Tan-sigmoid (tansig)	6	Saturating linear (satlin)	1	96.15	100	96.77
PNN	9	Radial basis function (RBF); (spread = 25)	112	Competitive (compet)		100	80	96.77