Douglas C. Montgomery Distinguished Lecture Series

I confess it’s a provocative title to get your attention. Given rich enough data, modern machine learning answers lots of questions by predicting the outcomes for cases that did not actually appear in the data. But if, for instance, you want to explore the financial viability of offering a hypothetical air taxi service in a city then there is no relevant data (at least not yet) from which to predict profits, much less to design an optimized version of the service. This is where computer simulation comes in. Computer simulation combines whatever useful data that exists with careful modeling to generate “fake data” that is relevant to the new scenario. Computer simulation is data analytics for systems that do not yet exist, and it lives at the boundary of industrial engineering and statistics, just like Prof. Montgomery.

In this talk I first describe and demonstrate “stochastic computer simulation” via an easy-to-understand (but real) example; no background is assumed. I then present two current challenges in the field: How to effectively exploit modern computing to optimize large-scale simulated systems, and how to adapt simulation to support real-time, data-driven decision making. The former topic is motivated by the availability of cheap, high-performance and ubiquitous parallel computing, and the latter by the emerging use of simulation as a “digital twin” to manage or control the real system it mimics. In both challenges I emphasize the new statistical issues that they imply using only very basic statistics to describe them. Thus, the talk should be easily understood by undergraduates and graduate students in engineering and statistics, and probably by their faculty as well.

Barry L. Nelson is the Walter P. Murphy Professor Emeritus of the Department of Industrial Engineering and Management Sciences at Northwestern. His research focus is the design and analysis of computer simulation experiments on models of discrete-event, stochastic systems, including methodology for simulation optimization, quantifying and reducing model risk, variance reduction, output analysis, metamodeling and multivariate input modeling. He has published numerous papers and three books, including Foundations and Methods of Stochastic Simulation: A First Course (second edition, Springer, 2021). Nelson is a Fellow of INFORMS and IISE.  

Further information can be found at Barry Nelson’s Home Page.

04/09/24

A Life in Monitoring Using Observed Data