Event Details

CENTRAL Exchange

Abstract:

The traditional use of simulation in Operational Research is in system design, either for studying systems or making tactical and strategic level decisions. This has been mirrored in the aims of simulation optimisation. In this case, once a decision has been made, the simulation will not be used again. It almost becomes irrelevant whether the simulation did a good job or not, as the stakeholders must make the best of the decision made. In the last 20 years, simulation (and simulation optimisation) has been used for operational decisions in Decision Support Systems. The model is reused repeatedly. So if the simulation is consistently producing poor predictions due to some systematic error, the results could be very costly. Thus, detecting this discrepancy becomes important. Due to natural variability in the system and simulation, this is not a trivial problem. It is complicated further by the decisions being heterogeneous and only getting one observation from the real world. This talk will present early work on a method for detecting systematic error in a simulation using a series of decisions.