Operations Research in Retail Analytics: An Interview with Dr. Barry Nelson
Professor Nelson teaches courses on computer simulation and statistical learning at Northwestern University in the Department of Industrial Engineering and Management Sciences. He was named McCormick teacher of the year in 1998 and 2007, was elected to the 2002, 2003 and 2007 ASG Faculty Honor Rolls, received the 2003 Northwestern Alumni Association Excellence in Teaching Award, was given the 2004 IIE Operations Research Division Award for Excellence in the Teaching of Operations Research, and was a Charles Deering McCormick Professor of Teaching Excellence at Northwestern.
Professor Nelson studies the design and analysis of computer simulation experiments, particularly issues of statistical efficiency (such as variance-reduction techniques), multivariate output analysis (such as multiple-comparison procedures and optimization via simulation), model risk (such as uncertainty quantification due to using estimated input models), multivariate input modeling (such as modeling and generation of nonstationary arrival processes), metamodeling (such as stochastic kriging) and simulation analytics (such as virtual statistics). His application areas include financial engineering, computer performance modeling, quality control, and manufacturing and transportation systems. He is also a Fellow of INFORMS and IIE.
Q1. Tell us a bit about your career?
A: My undergraduate degree was in mathematics from DePauw University, Indiana and I was interested in applied mathematics. So I decided to do a Masters in Industrial Engineering from Purdue where I was working on a large scale simulation model for Air Force. I found this problem interesting because I liked the challenges and the questions posed in simulation research and towards the very end of my Masters when I had job offers to decide on, I shifted to PhD. That was a time when simulation accelerated as an active field of research with Purdue as the center and I was at the right place at the right time.
Q2. You mentioned that you have a degree in mathematics, how much did that help in your research?
A: If we break simulation into three phases, there is programming, modelling and analysis. All three aspects are important for an effective simulation. Programming is vital when you have to simulate a very complex system in micro-level detail, and modelling is essential to represent the actual system accurately. Having done the first two things correctly, if we do not know how to analyze and interpret the results then all the effort is lost. Mathematics is useful for seeing simulation as both a realization of a stochastic process, and as a statistical experiment. This helps get the analysis right.
Q3. What are some of the present and future challenges and opportunities in simulation optimization research?
A: Exploiting massive parallel computing and incorporating (even defining in a useful way) stochastic constraints; these are my top two immediate challenges.
Q4. Simulation is used as a source for data generation rather than collecting data from real setup. How reliable do you think the data will be?
A: Validation of simulation has always been a problem in practice. Large scale models in various areas like supply chain, call centers and telecommunication systems are difficult to model making the validation even more difficult. That is why I think, quantification of model risk, rather than validation, is a promising research direction. We can do that to a limited extent now, with respect to input models, but there is much more to do.
Q5. What are the greatest advances you’ve seen recently in simulation optimization software?
A: The commercial software has been made easier and easier to use, and to apply to a broad class of simulation optimization problems. Stated differently, the focus has been on generality. You can always do better if you exploit specifics of the problem, but the mathematical programming folks have been better than we have been at exploiting particular problem structure in their commercial software.
Q6. Have you ever been in a low point during research? If so what advice would you give for the students facing similar situation?
A: Yes. When I was working on my research, there was a point at which we dropped the project we had worked on for about 3 months because it was not yielding any good direction. During that time, I focused on other work. That helped me to keep making progress after which I went back and was able to look at the original problem from a different perspective. I think it is essential to take a break, step back and turn your focus to something else when you hit a wall.
(Edited By: Karthick Gopalaswamy)