Options
Model-based Optimisation with Tree-structured Parzen Estimation for Container Terminals
Publikationstyp
Conference Paper
Date Issued
2019-09-19
Sprache
English
Author(s)
Institut
TORE-URI
Start Page
489
End Page
498
Citation
Simulation in Produktion und Logistik (ASIM 2019)
Contribution to Conference
Publisher
Wissenschaftliche Scripten
Traditionally discrete event simulation serves as a tool for estimating the impact of managerial decisions at container terminals. For complex situations, simulation-based optimisation can be employed to lower the amount of executed simulation experiments and therefore reduce the time to obtain results. The search is guided by the already run experiments. This paper examines the applicability of the Tree-structured Parzen Estimator for optimizing simulation models of container terminals. The reasoning is based both on prior literature in the field of simulation and machine learning, as well as a numerical study. Within the experiments, the performance of the Tree-structured Parzen Estimator is compared with Simulated Annealing and Random Search. The Tree-structured Parzen Estimator displays a more explorative and robust search behaviour than Simulated Annealing.
Subjects
Container Terminal
Simulation-based optimisation
Hyperparameter Optimization
DDC Class
330: Wirtschaft
Version History
You are currently viewing version 1 of the item.