Genetic_Programming_Theory_and_Practice_XIII

Prime-Time: Symbolic Regression Takes Its Place in the Real World 259

Ta b l e 2 Parameter design for all modeling iterations with FluTE and obtained variable

importance for the cumulative clinical attack rate

Parameter RUN 1 RUN 2 RUN 3 RUN 4 RUN 5

Region Seattle Seattle LA County Seattle LA County

Travel allowed? Yes/no () Yes/no () Yes/no () yes yes/no ()

R 0 1.1–2.4 (++)1.1–2.4 (++)1.1–2.4 (++)1.1–2.4 (++)1.1–2.4 (++)

Infected seeds 0–5000 (+*)0–1024 (+*)0–1024 (+*)0–1024 (+*)0–1024 ( +*)

Seeded daily? Yes/no (++) Yes/no (+) No Yes/no (-) No

Ascertainment 0–90 % (+) 80 %

Ascertainment delay 1–5d () 1d

Response threshold 0–5 % (+) instant

Response delay 0–30d () instant

Vaccination coverage 0–90 % () 0–90% (++)

VEsusceptibility 0–66 % () 0–66 % (+)

VEinfectiousness 0–66 % () 0–66 % ()

VEsymptoms 0–66 % () 0–66 % ()

Scenarios 200 200 50 800 200

Repetitions 20 20 10 20 20

Legend: ++ very important, + important,almost no impact, * only small values, VE:

Vaccine efficacy

interactive profiler of a simulator meta-model provides unique selling points for
several reasons:

• It is critical to understand simulator inputs that impact the Key Performance
  Indicators, and SR does just that.


• It is critical to get reliable predictions for predictions of simulator outputs as
  well and the trustability of the predictions. Robust ensemble-based symbolic
  regression does just that.


• The only way to facilitate data-driven decisions is to empower the decision maker
  with simple to use tools to explore what-of scenarios and be best prepared for
  whatever is coming.

References

Andradóttir S, Chiu W, Goldsman D, Lee M, Tsui K, Sander B, Fisman D, Nizam A (2011)
Reactive strategies for containing developing outbreaks of pandemic influenza. BMC Public
Health 11(Suppl 1):S1
Chao D, Halloran M, Obenchain V, Longini I (2010) FluTE, a publicly available stochastic
influenza epidemic simulation model. PLoS Comput Biol 6(1):e1000,656
Crombecq K (2011) Surrogate modelling of computer experiments with sequential experimental
design. Ph.D. thesis, University of Antwerp, Antwerp
Crombecq K, Dhaene T (2010) Generating sequential space-filling designs using genetic algo-
rithms and monte carlo methods. In: Simulated evolution and learning. Lecture notes in
computer science, vol 6457. Springer, Berlin, pp 80–84