Dynamic Queueing Networks: Simulation, Estimation and Prediction
Student: Anthony Ebert
When: Wednesday, 1 March 2017 3:00 PM-4:00 PM
Where: GP-Z Block, Level 4, Room 413
- Prof Kerrie Mengersen (Principal)
- Dr Paul Wu
- Prof Fabrizio Ruggeri
- Kerrie Mengersen (Chair)
- Paul Wu
- Fabrizio Ruggeri
- Christopher Drovandi
- Clinton Fookes
Important real-world systems such as airport terminals, manufacturing processes and hospitals are modeled with networks of queues. To estimate parameters, restrictive assumptions are placed on these models. For instance arrival and service distributions are assumed to be time-invariant. Violating this assumption are so-called dynamic queueing networks which are more realistic but do not allow for likelihood-based parameter estimation.
If we could simulate the model quickly then we could use likelihood-free techniques to estimate parameters. If we could estimate parameters then we could predict performance. If we could predict performance then we could use the model for planning and even real-time decision making.
Recent developments in Lindley-type Recursive Relations (LRR) and Approximate Bayesian Computation (ABC) offer improvements in simulation speed and likelihood-free parameter estimation respectively. We aim to extend, with methodology and software, these developments to dynamic queueing networks, with contributions to simulation, estimation and prediction.