By Siddharth Vodnala, Voiland College of Engineering and Architecture

Like many of us, Yunshu Du enjoys working out at Washington State University’s Student Recreation Center (SRC). She gets annoyed, though, when she has to search for an open treadmill or wait for her fellow students to finish using the weight training equipment.

Du, a Ph.D. student in computer science, decided to come up with a solution to the problem.

Last year, she was taking classes on data science and machine learning and began wondering if she could apply the techniques she was learning in those classes to the SRC. She approached her professors, Assefaw Gebremedhin and Matthew Taylor from WSU’s School of Electrical Engineering and Computer Science, who helped her launch the project.

“I began to wonder if there was a smart way to consistently avoid crowds,” she said.

Reviewing data from student ID cards that are swiped when students use the SRC and then applying machine learning techniques, the researchers showed how data can be harnessed to find patterns in and predict recreation center usage. The research can help ease congestion in such facilities and help managers provide a better experience to students.

A paper detailing their work appears in IEEE Transactions on Knowledge and Data Engineering.

Analyzing the data, the researchers found that Mondays were the busiest days of the week at the SRC. Weekday afternoons were busy, while weekend mornings saw higher usage. Spring semester saw more overall attendance than during the fall, and men worked out more often than women. Freshmen also used the facility more than upperclassmen or graduate students.

Closeup of Assefaw Gebremedhin.
Assefaw Gebremedhin

The work showed that card swipes can be a rich source of data that can help improve student services, said Gebremedhin.

The project was the first in which Du had the chance to use real world data.

“I had to do a lot of pre-processing, cleaning of the data and had to figure out the best way to visualize it,” she said.

The researchers then used statistical and machine learning techniques to figure out the best time to visit the SRC. They also built a web application that could be accessed by SRC managers.

The page helps the managers decide how much staff to have on duty on a given day. Knowing the busiest times, student employees learned the best times to restock towels and check on equipment. Some of the employees who also worked out at the gym used the web app to plan their exercise times.

“This is another example of how data science is making our lives simpler and better,” said Gebremedhin.

The researchers plan to make the tool available to all students in the future. They also plan to add a feature for recommendations that are personalized to the user’s lifestyle. For example, workout prompts could be sent specifically to seniors and graduate students, since they workout less frequently than other students.

Two students working out at the SRC.
Two students working out at the SRC

Apart from showcasing the data science process, this project can also be a great learning tool for computer science students, said Gebremedhin.

Du received a university-wide graduate student award for this project.

This project is supported by grants from the National Science Foundation (IIS-1734558 and IIS-1553528); the National Aeronautics and Space Administration (NNX16CD07C); and the U.S. Department of Agriculture (2014-67021-22174).