The design and analysis of modern discrete-event, stochastic simulation has been greatly influenced by its heritage in queueing theory; and the strength of queueing theory is in deriving long-run performance measures for stationary service systems. Data analytics, on the other hand, stresses uncovering conditional relationships and making predictions. The success of data analytics in business and industry will lead simulation users to expect this sort of fine-grained, conditional analysis from their simulations, and if they cannot obtain it they may conclude that simulation is irrelevant. The proposed research provides a foundation and proof-of-concept first steps toward a data analytics treatment of dynamic, stochastic simulation. The proposal is not concerned with the use of “big data” to drive or parameterize a simulation, nor does it investigate the use of simulation to assist in the analysis of “big data.” Instead, it addresses simulation in its traditional role for system design, and its less traditional role for system control. The revolution is to treat stochastic simulation as data analytics for systems that do not yet exist, rather than as approximate queueing theory, and to extend the reach of simulation beyond parameter estimation and system optimization, and toward performance prediction and uncovering the key drivers of system behavior. Three core topics are proposed: simulation analytics as a precursor to system control, simulation analytics for comprehensive comparisons, and simulation analytics via dynamic metamodels. The focus is on a post-simulation analysis that is facilitated by retaining the simulated sample paths, sample paths that may have been generated by a simulation experiment designed to achieve a specific narrow objective (e.g., optimization). The research seeks to facilitate investigating anticipated and unanticipated questions, discovering relationships, understanding system behavior, and sometimes suggesting additional simulation experiments that might be fruitful. Collaboration with our GOALI partner SAS Institute will insure that the research is relevant and the results implemented.
|Effective start/end date||9/1/15 → 8/31/19|
- National Science Foundation (CMMI-1537060)
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.