A Heuristic Detector Generation Algorithm for Negative Selection Algorithm with Hamming Distance Partial Matching Rule - Artificial Immune Systems - page 236

G N N

Table 3. Comparisons on between h-NSA and t-NSA when fixing and varying

M S R

N Max G Min G AVG G

STDDEV

R M M M

h-NSA 16120 15167 15675.7 365.22

50

t-NSA 28545 22280 24527.7 1792.28

h-NSA 31513 30152 30879.1 523.31

100

t-NSA 54513 46722 49757.6 2552.35

h-NSA 47259 45413 46127.4 646.26

150

t-NSA 85156 72262 77705.6 3711.31

h-NSA 62433 60703 61734.7 545.77

200

t-NSA 109638 94355 101503.9 4410.65

h-NSA 79305 76100 77314.0 1008.22

250

t-NSA 139078 121987 126130.4 5260.17

h-NSA 94836 91775 92897.3 1205.18

300

t-NSA 163775 148562 155445.2 5053.12

h-NSA 108330 106551 107743.4 600.72

350

t-NSA 187053 175327 182479.2 3808.99

h-NSA 125041 122900 123915.1 656.90

400

t-NSA 215162 197034 204130.6 5960.36

h-NSA 141056 138041 139746.7 1108.97

450

t-NSA 242589 222226 232088.5 6304.61

h-NSA 156130 153357 154610.9 953.22

500

t-NSA 260032 246993 252299.7 4340.26

shown in Table 3 and Table 4. For convenience, the experimental results are also
shown in Fig. 3 and Fig. 4.
From Table 3 and Fig. 3, it can be observed that when

N is fixed, the

C

in both

S

h-NSA and t-NSA almost have no notable changes with the increment of N . How-

ever, when N is fixed, it's much more difficult to generate the same number of
matured detectors in t-NSA than in h-NSA.
From Table 4 and Fig. 4, it can be observed that the detection ratio

R

D of both h-

R

NSA and t-NSA increase with the increment of N . But also obviously, the detection

ratio D of h-NSA is much higher than that of t-NSA for the same number of ma-
tured detectors.
From Fig. 3 and Fig. 4, It is also known that the standard deviations of both

R

C

and D in h-NSA are lower than those in t-NSA. So, the h-NSA is more stable than

R

t-NSA.