Home / Science / Where some see numbers, a USC scholar spots a story worth telling

Where some see numbers, a USC scholar spots a story worth telling


Where some see numbers, a USC scholar spots a story worth telling

Numbers ninja Bradley Rava uses statistics to reveal untold tales buried in huge sets of data

BY Eric Lindberg

In the multibillion-dollar world of horse racing, gamblers are always looking for a competitive edge.

From insider tips on a horse’s health to the track’s soil condition, there are countless variables to consider before placing a bet with hopes of walking to the pay window.

Sorting through that complex tangle of information is overwhelming to most people, but for Bradley Rava, it’s an exciting challenge. The first-year PhD student in statistics at the USC Marshall School of Business revels in unearthing valuable secrets hidden in massive data sets.

He recently crunched the numbers on more than 4,000 horse races in the United Kingdom to determine whether being assigned to a specific stall at the starting line can influence the chances of winning. His research, part of a class project during a fellowship at Yale University, revealed that three of the four tracks he analyzed had “draw bias.” That’s industry lingo for advantageous or unfavorable starting positions.

Where others might only see reams of numbers, Rava sees a story waiting to be told. The 23-year-old, who completed his undergraduate degree in applied and computational mathematics at USC in 2016, draws inspiration from finding patterns that reveal something about the world around him.

He has used his analytic skills to study the likelihood that a cybercriminal might target a specific company. He is currently considering how to refine formulas used by companies like ESPN to predict the winner of a sporting event as the game unfolds.

“There is a lot of math, and the math is incredibly important, but there is also a huge element of creativity,” Rava said. “That is what separates people who might be good at analyzing data from those who can use data to tell a story and make some kind of meaningful impact on society.”

His ability to pull valuable information from big data sets caught the eye of the National Science Foundation, which granted him a three-year research fellowship that covers his tuition at USC and provides other perks, like conference funding and access to a powerful quantum computer.

No limits

He is still narrowing down his interests, which include machine learning, network analysis and high-dimensional statistics. But Rava said he was drawn to applied math and statistics in general because of their usefulness in virtually every field, from biology and neuroscience to economics and business.

“Everybody needs math,” he said. “It’s an underlying language that we can use to describe processes around us. It’s cool to be able to jump into a new field and find something useful using that language.”

That ability recently came in handy during the Citadel Data Open, a regional “datathon” held at Caltech that pitted teams of coders, programmers and data scientists against the clock in a race to analyze a large data set featuring transportation information from New York.

Rava and his teammates, fellow USC Marshall PhD student Michael Huang and collaborators from UCLA (Baichuan Yuan) and the University of California, Davis (Jiaping Zhang), spent hours hunched over their laptops in a fervent frenzy of coding, barely pausing to wolf down a quick lunch.

“I get fairly competitive with these things, so it was pretty intense,” Rava said. “Luckily, we came up with something the judges liked, and we ended up winning. It was the best seven hours of my life.”

The judges cited their rigorous and logical approach to analyzing how Uber is gaining the edge over taxis in the city’s outer boroughs, awarding them with $20,000 in cash and an invitation to the national competition in New York. They didn’t win that event, but they’re already planning strategies for the next showdown.

Up to code

A native of El Segundo, Rava can’t point to a specific moment in his life that sparked his interest in math. He admits he didn’t grow up with an innate gift for analyzing numbers, but always enjoyed the struggle to solve a problem and reach a definitive answer. “To be good at anything,” he said, “you just need to work hard at it.”

He figured out quickly as an undergrad that he was more interested in applied math rather than what he termed pure math.

Pure mathematicians, he explained, find enjoyment in creating and refining analytic tools and theories with little consideration of how they might be used.

“That’s super important and valuable,” Rava said, “but I look at these tools and think, these are awesome because we can use them to understand the world around us and help shape policy and answer important questions.”

After finishing his undergraduate studies at USC, he received a one-year Emerging Scholars Initiative Fellowship to attend Yale University, allowing him to take courses on various statistical strategies that bolstered his interest in real-world applications of math.

One class focused on topological data analysis, which expresses large data sets visually. This allows researchers to identify patterns and connections that could otherwise remain hidden. Rava considered how researchers might use that technique to study the looping nature of artery trees in the human brain, drawing connections from those delicate structures to the risk of stroke.

Both of his advisers at USC Marshall’s Department of Data Sciences and Operations — Gareth James and Xin Tong — specialize in similar analytic techniques involving high-dimensional statistics and machine learning. Although he is still settling into his first year in the PhD program, Rava envisions pursuing a research career involving those approaches.

“For now, I’m exploring a lot of problems, being exposed to a lot of tools I didn’t know about and learning as much as I can,” he said.

source: University of Southern California

About Mohammad Daeizadeh

Leave a Reply

Your email address will not be published. Required fields are marked *