Existing models of multi-server queues with system transience and non-standard assumptions are either too complex or restricted in their assumptions to be used broadly in practice. This paper proposes using data analytics, combining computer simulation to generate the data and an advanced non-linear regression technique called the Alternating Conditional Expectation (ACE) to construct a set of easy-to-use equations to predict the performance of queues with a scheduled start and end time. Our results show that the equations can accurately predict the queue performance as a function of the number of servers, mean arrival load, session length and service time variability. To further facilitate its use in practice, the equations are developed into an open-source online tool accessible at http://singlequeuesystemstool.com/. The proposed procedure of data analytics can be used to model other more complex systems.
Data Analytics for Queues, Simulation, Nonlinear regression, Alternating Conditional Expectation
Operations and Supply Chain Management | Operations Research, Systems Engineering and Industrial Engineering
Computers and Operations Research
YANG, Kum Khiong; TUGBA, Cayirli; and LOW, Mei Wan.
Predicting the Performance of Queues: A Data Analytic Approach. (2016). Computers and Operations Research. 76, 33-42. Research Collection Lee Kong Chian School Of Business.
Available at: http://ink.library.smu.edu.sg/lkcsb_research/4944