Validation of a Novel Traditional Chinese Medicine Pulse Diagnostic Model Using an Artificial Neural Network

Table 3

Comparison of the specificity, sensitivity, and accuracy of the three back-propagation training algorithms with different numbers of hidden neurons (𝑁=260).

Algorithm	Number of hidden neurons	Specificity; (%)	Sensitivity; (%)	Accuracy; (%)
Bayesian regularization	10	69.96	76.57	73.50
	15	73.46	84.49	79.27
	20	73.20	84.80	79.30
	25	69.33	84.80	77.51
Levenberg-Marquardt Algorithm	10	68.29	86.75	78.06
	15	63.17	90.88	77.83
	20	63.67	88.04	76.56
	25	62.93	87.40	75.87
Resilient backpropagation	10	72.19	83.91	78.41
	15	63.67	91.33	78.28
	20	66.61	91.30	79.67
	25	65.85	90.22	78.74