Fall 2005 Seminar Series

DATE: Thursday, September 15, 2005
LOCATION: E40-298
TIME: 4:15pm
Reception immediately following in the Philip M. Morse Reading Room, E40-106

SPEAKER:
Stephen E. Chick
Associate Professor
Technology Management Area
INSEAD

TITLE
A Bandit Approach to the Economics of System Selection with Simulation

ABSTRACT
Managers that select the operating characteristics of manufacturing, supply chain, or service delivery systems often aid their decision-making with stochastic or discrete event simulation. The simulation selection problem presents a tension between the desire to reduce the risk of an incorrect selection by simulating more to better learn the mean net present value (NPV) of each alternative, versus a need to implement early and avoid prolonged simulation analysis costs and discounting costs associated with a delayed system implementation.

 

This paper presents a new approach to the simulation ranking and selection problem that maximizes the NPV of decisions made with stochastic simulation. We pose the problem formally as a stoppable bandit process under relatively general conditions, show that a Gittins index can be used to indicate which system to simulate or implement, and provide an asymptotic approximation for the index that is appropriate in a simulation context with normally distributed output with known but potentially different variances for each system. We conclude with a discussion that relaxes some of those assumptions.

 

Joint work with Noah Gans.