Genetic_Programming_Theory_and_Practice_XIII

258 S. Stijven et al.

a

1.2 1.4 1.6 1.8 2.0 2.2 2.4

0.0

0.1

0.2

0.3

0.4

0.5

R 0

Attack Rate

0 200 400 600 8001000

0.0

0.1

0.2

0.3

0.4

0.5

Infected people seeded

Attack Rate

b

1.2 1.4 1.6 1.8 2.0 2.2 2.4

0

20

40

60

80

100

120

140

R 0

Day Epidemic Peak

0 200 400 600 8001000

0

20

40

60

80

100

120

140

Infected people seeded

Day Epidemic Peak

Fig. 11Response prediction plots for Seattle and LA County. [Adapted from Willem et al. ( 2014 )]
Response prediction plots of the high-quality surrogate models obtained with SR for the cumulative
clinical attack rate (a) and the day of the epidemic peak (b) in Seattle (black) or LA county (gray).
Predictions for R 0 assume a fixed number of infected seeds, indicated by thedotin the panel on
theright, and vice versa

et al. 2005 ). The clear visualization of herd immunity with the surrogate models
emphasizes the usefulness of our approach since it is hard to observe this effect
directly from the numerous individual simulation results. An interactive version of
this plot is available atwww.idm.uantwerpen.be.

4.4 Summary of Modern Complex System Analysis

Symbolic Regression combined with intelligent design of experiments has proven
to be an indispensable tool in understanding complicated simulation models and
reducing them to practice. The process of replacing a black-box simulator by an