An Automatic Source Code Vulnerability Detection Approach Based on KELM

Table 3

Effect of different number of neurons on the precision of ELM.

Neuron number	FPR (%)	TPR (%)	P (%)	F1 (%)	Training time (s)
250	7.5	58.2	75.4	65.7	0.92
500	6.6	66.1	80.0	72.4	2.52
1000	6.4	72.8	81.9	77.1	7.75
3000	5.4	80.5	85.5	82.9	49.11
5000	4.4	83.6	88.3	85.9	128.72
10000	4.4	85.9	88.6	87.2	496.02
12000	4.3	85.9	88.9	87.4	640.68
15000	4.4	86.6	88.6	87.6	1143.70
20000	4.6	86.7	88.2	87.4	2201.60