Please Note: This seminar will be given online.
Selection of the most probable best
In many business applications, simulation is the primary decision-making tool for a complex stochastic system, where an analytical expression of the problem is unavailable. Often, parameters of these simulators are unknown and must be estimated from data. When plug-in estimates of the parameters are adopted, there is a risk of making a suboptimal decision due to the estimation error in the parameter values.
Under this type of model risk, this talk discusses a new decision-making framework in the context of simulation optimization, the most probable best (MPB), is introduced in this talk. The MPB is defined as the solution whose posterior probability of being optimal is the largest given the data when the parameters’ estimation error is modeled with a posterior distribution. Some saliant theoretical properties of the MPB will be discussed including its strong consistency to the optimum under the true parameter as the data size increases. In the second half of the talk, efficient sequential sampling algorithms to find the MPB will be introduced and their asymptotic optimality (in efficiency) will be discussed. To demonstrate business insights the MPB formulation provides, a product portfolio optimization problem, where consumer utility parameters are estimated from conjoint survey data will be presented.