By Gary Goettling
Analytics is attracting a great deal of attention in the business world these days, and no one knows that better than Joel Sokol, the Fouts Family Associate Professor at the H. Milton Stewart School of Industrial and Systems Engineering (ISyE) at Georgia Tech.
Sokol also serves as director of Georgia Tech’s new interdisciplinary Master of Science degree in Analytics, which graduated its first class this past August.
“We couldn’t advertise the degree until it was formally approved by the Board of Regents, and that happened in late May last year,” says Sokol. “So we missed the entire application season, which usually happens in the fall and spring. We were hoping we could scrape up 20 people to put this class together, but we got about 80 applications in just a few weeks. We accepted 44 outstanding applicants. All but three students enrolled, with a couple students deferring their admission, so we started with 39.”
Word got out. The current class, which started in August 2015, drew more than 400 applicants, from which a class of 47 was selected.
These numbers reveal a growing interest in analytics among businesses and organizations as a way to analyze and interpret the data they acquire. At ISyE, where analytics has been central to its educational mission for many years, this interest is reflected not only in the number of applications to the new master’s program, but also in analytics research into contemporary problems and a rigorous interdisciplinary curriculum that includes Senior Design projects for undergraduate seniors.
The basic definition of analytics is the extraction of meaningful information from data. At ISyE, this definition goes a key step further.
“For us, analytics is not only the techniques to process data and extract information from data or even the knowledge obtained from data,” says Martin Savelsbergh, James C. Edenfield chair and professor. “It is also how you can use that information or knowledge to improve business processes or make better decisions."
“Analyzing data per se is not what people are after,” he adds. “In the end, you want to use what you learn from data to be better at something, and being better at something usually means that you make better decisions.”
Savelsbergh points out that the ability to collect, store, and manipulate data has grown exponentially over the years, along with sophisticated techniques and algorithms for analyzing it. But more data in and of itself isn’t necessarily the right objective.
There can be data overload, he says, citing an actual example of a trucking company that outfitted its vehicles with GPS and two-way communication so as to better keep track of its fleet.
“Every five minutes the company gets updated information from each truck regarding the time it is estimated to arrive at the distribution center. At one point, a signal is received that says a particular truck is expected in one and a half hours. Five minutes later the signal says the truck is expected in one hour and 40 minutes. Five minutes after that the signal says the truck will arrive in an hour and 45 minutes — and so on.
“Is this really very useful information?” Savelsbergh asks. “What are you going to do with that information? Certainly the people at the trucking company don’t know what to do with it.”
While conceding that more information is generally better than less information, one must be careful to avoid collecting data simply for its own sake.
“You want to process, analyze, and understand the data — maybe understand trends — but whatever it is you’re looking for in the data, you want to be sure it helps you make better decisions about something.”
The practical, decision-enabling orientation of analytics at ISyE is evident in the new master’s program as well, which includes an applied analytics practicum at the end of the one-year program.
Graduate analytics degree programs are relatively uncommon in the U.S., and the majority of those that do exist are part of a particular college or school.
“Ours is one of the handful that’s interdisciplinary,” says Sokol. “It’s a joint program among the College of Engineering, the College of Computing, and the Scheller College of Business.
“Students get an interdisciplinary core that covers a full range of analytics topics, and then they pick a track to get a deeper specialization. Each of the tracks is aligned with one of the three units. We have an analytical tools track that includes additional statistics, and ML and OR predictive and decision modeling material. Students who opt for the business analytics track get a deeper understanding of the practice of developing and executing analytics projects within businesses. In the computational data analytics track, students get additional depth in acquiring, managing, analyzing, and visualizing data.”
Another unusual feature is that half or more of the 10 courses that students take are electives, which allows them to tailor their degree to fit their personal career interest.
“They may take courses in specific areas so they can perform the right kind of analysis for whatever industry they want to go into,” Sokol elaborates. “For example, if they want to do analytics in the hospitality industry, they might take electives on pricing and revenue management, Web search and text mining, and optimization so they could capture and analyze data such as from TripAdvisor and, then use the results to suggest improved pricing policies.”
The broad applicability of analytics is reflected in the diverse backgrounds of the program’s applicants.
“The majority of them come in with degrees in business, engineering, math, statistics, or computer science, but we also get people with degrees in psychology, anthropology, astrophysics, linguistics, religion — a whole range of backgrounds,” Sokol notes.
A variety of job experience is represented as well, with about 60 percent of applicants having had some previous employment.
Their resumes run the gamut from just a few years of post-bachelor’s degree work experience to positions as lead product engineers and corporate vice presidents. One applicant had spent the past few years in the U.S. Navy aboard a nuclear submarine, according to Sokol.
When an Atlanta-area hospital wanted to cut wait time and provide more accurate wait-time estimates for its emergency room patients, it sought help from a group of Georgia Tech industrial engineering undergraduate students who took on the assignment as their Senior Design project.
Senior projects are an integral part of the undergraduate curriculum, says Sokol, who supervises the program each fall. “At the end of their time at Georgia Tech, students form groups and carry out real industrial engineering projects for companies and organizations that need their help. A lot of what they do involves analytics, but at the undergraduate level.”
Back at the hospital, the team of six students observed emergency room operations over several months. They collected and analyzed data on patient arrivals and conditions as well as the amount of time taken to be seen by nurses, doctors, and other hospital staff. The students then used statistical techniques to model the emergency room system in a simulated environment, and devised process-improvement recommendations that would reduce wait time without changing the quality or quantity of care. They also developed a real-time simulation tool that helps the hospital give entering patients a more accurate estimate of their wait time.
Other examples of senior capstone projects reveal the wide applicability of analytics and include delivery routing and logistics for supply chain design, pricing for hotels and parking, race strategy for a motor sports team, and the timing of trains and railcar sequencing.
In addition, student teams have worked with the Centers for Disease Control and Prevention on various aspects of their response to the Ebola outbreak, and improvements to the organ transplant system.
Data analytics research conducted at ISyE explores innovative new methodologies and techniques for analyzing data across a spectrum of applications, from energy and finance to supply chains and sports.
But no other area affects the quality of daily life for more people than their health.
Arriving at the best medical and health care decisions relies heavily on data, says Nicoleta Serban, Coca-Cola Associate Professor of ISyE. “We are interested in finding ways to capture and analyze data to optimize the decision-making process in health care.”
Her work extends into the arena of public health as well, where policymakers need data-driven conclusions to help them make effective problem-solving decisions.
Serban is co-founder and co-leader, along with Harold R. and Mary Anne Nash Professor Julie Swann, of the Health Analytics Group. Its mission is to provide a foundation for better medical decisions by applying mathematical and computational modeling techniques to health services research data and health economics data.
One of the challenges of health care analytics is that it may deal with so-called Big Data — huge data sets measured in terabytes and exabytes — but not always.
“The quality of data is a more important consideration than the volume of data,” Serban says. “The key term is ‘decision-making’ — that data is captured and analyzed for the purpose of making better decisions. Sometimes this involves Big Data, and sometimes it involves very little data.”
The Health Analytics Group’s wide-ranging research interests address both traditional and emerging health analytics models, including:
One example of a specific research initiative is the group’s ongoing study of pediatric asthma.
Asthma — the second most common reason for pediatric emergency room visits in Georgia — impairs quality of life and contributes significantly to health care costs, particularly for emergency room visits and hospitalizations, many of which are preventable. These costs are especially burdensome to children from low-income households.
“Our immediate objective is to describe underlying asthma care pathways for children in the Medicaid program,” Serban explains. “For each pathway, we evaluate utilization and costs to suggest potential policy and network interventions.”
Designing interventions with the greatest impact on patients with limited resources begins with the creation of an asthma care baseline.
“We want to quantify a set of measures around pediatric asthma for the Medicaid population,” she notes. “Our initial baseline includes things related to outcomes and costs, and for geographical areas and subpopulations within the state of Georgia.”
Baseline data would include many of the complicating factors in treating pediatric asthma such as age, severity of the condition, and environment.
In addition, there are different levels of asthma care to consider, from doing nothing to obtaining care from a primary care physician or asthma specialist, or visiting the emergency room.
“Using retrospective Medicaid claims data, our research spans multiple directions,” she says. “In addition to the set of baseline measures for asthma care, we’re interested in linking access to outcomes, and identifying trends in care utilization and cost.
“Ultimately, our goal is to design policy and network interventions to improve health outcomes and access for people with limited resources.”
Industrial engineers are problem solvers, which is why analytics is considered an engineering discipline.
“We’re not concerned with building or designing physical objects,” says Savelsbergh. “We’re interested in processes and in finding ways to improve the performance of businesses and organizations.” Thus, analytics is a natural fit.
“There are still a lot of people who are trying to understand what is meant by analytics,” he continues, “and this gives us an opportunity to interact with organizations either in government or private industry to not only talk about what we do in analytics, but to emphasize our belief that its goal is to improve decision-making.
“We have been doing analytics for a long time -- and we're very good at it."