A State-Based Sensitivity Analysis for Distinguishing the Global Importance of Predictor Variables in Artificial Neural Networks

Table 4

Resultant mean values for input ( to ) and output () variables in the hypothetical database derived in state-based sensitivity analysis (SBSA) and local sensitivity analysis (local SA), for instance, when the mean value of lies within a distinct “state” along its distribution range.

state	value	SBSA					Local SA
		1	—	—	—	—	—	—	—	—	—	—	—
2	—	—	—	—	—	—	—	—	—	—	—
3	−1.83	−0.55	0.12	−0.44	0.04	−1.50	0.01	0.16	−0.27	0.03	−1.29
4	−1.37	−0.01	0.96	−0.20	1.03	−1.10	0.01	0.16	−0.27	0.03	−0.95
5	−0.91	−2.16	−1.56	−1.14	−0.45	−1.61	0.01	0.16	−0.27	0.03	−0.62
6	−0.46	−0.28	0.03	0.03	−0.70	−0.28	0.01	0.16	−0.27	0.03	−0.30
7	0.00	0.44	−0.35	−0.75	0.95	−0.09	0.01	0.16	−0.27	0.03	0.04
8	0.46	0.59	1.04	0.20	−0.04	0.75	0.01	0.16	−0.27	0.03	0.37
9	0.91	1.34	−0.72	−0.44	−0.70	1.30	0.01	0.16	−0.27	0.03	0.69
10	—	—	—	—	—	—	—	—	—	—	—
11	1.83	−0.28	−1.28	0.49	−0.20	0.83	0.01	0.16	−0.27	0.03	1.36
12	—	—	—	—	—	—	—	—	—	—	—