A typical approach for computing an optimal strategy for non-pharmaceutical interventions during an influenza outbreak is based on statistical ANOVA. In this study, for the first time, we propose to use bi-objective mixed integer linear programming. Our approach employs an existing agent-based simulation model and statistical design of experiments presented in Martinez and Das (2014) to generate the required data. Our computational results show that for an influenza outbreak with maximum attack rate of 33%, if we use the proposed optimal policy then the attack rate reduces to 0.6% with 85% accuracy. Our result significantly outperforms the existing result in the literature that estimates the influenza attack rate to be 1.83% under the proposed optimal policy.
ORHC_2017_137, University of South Florida, 4202 E. Fowler Ave. ENB118 Tampa, FL- 33620, October 2017
View Best subset selection of factors affecting influenza spread using bi-objective optimization