Research output: Contribution in Book/Report/Proceedings - With ISBN/ISSN › Conference contribution/Paper › peer-review
Research output: Contribution in Book/Report/Proceedings - With ISBN/ISSN › Conference contribution/Paper › peer-review
}
TY - GEN
T1 - Genetic algorithms for calibrating airline revenue management simulations
AU - Vock, Sebastian
AU - Enz, Steffen
AU - Cleophas, Catherine
PY - 2015/1/23
Y1 - 2015/1/23
N2 - Revenue management (RM) theory and practice frequently rely on simulation modeling. Simulations are employed to evaluate new methods and algorithms, to support decisions under uncertainty and complexity, and to train RM analysts. To be useful in practice, simulations have to be validated. To enable this, they are calibrated: model parameters are adjusted to create empirically valid results. This paper presents two novel approaches, in which genetic algorithms (GA) contribute to calibrating RM simulations. The GA emulate analyst influences and iteratively adjust demand parameters. In the first case, GA directly model analysts, setting influences and learning from the resulting performance. In the second case, a GA adjusts demand input parameters, aiming for the best fit between emergent simulation results and empirical revenue management indicators. We present promising numerical results for both approaches. In discussing these results, we also take a broader view on calibrating agent-based simulations.
AB - Revenue management (RM) theory and practice frequently rely on simulation modeling. Simulations are employed to evaluate new methods and algorithms, to support decisions under uncertainty and complexity, and to train RM analysts. To be useful in practice, simulations have to be validated. To enable this, they are calibrated: model parameters are adjusted to create empirically valid results. This paper presents two novel approaches, in which genetic algorithms (GA) contribute to calibrating RM simulations. The GA emulate analyst influences and iteratively adjust demand parameters. In the first case, GA directly model analysts, setting influences and learning from the resulting performance. In the second case, a GA adjusts demand input parameters, aiming for the best fit between emergent simulation results and empirical revenue management indicators. We present promising numerical results for both approaches. In discussing these results, we also take a broader view on calibrating agent-based simulations.
U2 - 10.1109/WSC.2014.7019894
DO - 10.1109/WSC.2014.7019894
M3 - Conference contribution/Paper
C2 - 25246403
SN - 9781479974863
SP - 264
EP - 275
BT - Proceedings - Winter Simulation Conference 2014
PB - Institute of Electrical and Electronics Engineers Inc.
ER -