When Rachel Price enrolled in Physics 5680: Big Data Analytics in Physics, she expected a challenging elective. What she didn’t expect was a new career goal.
As an astrophysics major at The Ohio State University, Price was comfortable with numbers—but machine learning, a subfield of artificial intelligence, was entirely new territory.
Physics 5680 is designed with students like her in mind.
“This course is an introduction to machine learning and advanced algorithms, with an emphasis on practical applications using publicly available datasets,” said instructor Richard Hughes, professor of physics. “The goal is to prepare students to be ready for future work either in academia or industry. Data analysis plays a huge role in both research and commercial settings, and I want students to hit the ground running.”
The course introduced students to data science through hands-on projects. While the content leaned toward physics-based problems, Price took a different approach for her final project—one rooted in her personal interests.
“I wanted to do something fun—something I could actually talk to my friends about and share with others who could quickly understand,” Price said.
That idea turned into a fantasy basketball simulator powered by the Ohio Supercomputer Center (OSC). With access to OSC’s classroom tools through Open OnDemand, Price trained a data science tool called the Random Forest Regressor model on five decades of NBA statistics—including draft and per-game performance data—to predict player output and generate an optimized fantasy lineup for the upcoming season.
Hughes said Price's choice immediately stood out in the class.
“Although this is a physics course, I don’t restrict final projects to science. I encourage students to pick topics that interest them—I want them to have fun with the knowledge they’ve gained,” he said. “Rachel’s project combined skill and creativity. She picked a specific problem, tested models rigorously, and made testable predictions. That’s exactly what I hope to see from students.”
Price initially tested three machine learning approaches: Linear Regression, Time-Series Forecasting, and Random Forest Regressor. The latter proved the most effective, striking the right balance between accuracy and flexibility. Random Forest Regressor models work by building many decision trees and averaging their results, which helps handle noisy or unpredictable data—like sports statistics—more effectively than simpler models.
Price’s final model predicted key performance metrics—points, assists, steals, rebounds, and more. It then optimized a fantasy roster of 10 players, following standard league constraints like salary caps, position requirements, and team diversity.
“Running that kind of data on a personal machine would’ve taken hours upon hours,” she said. “With OSC, I had results in minutes.”
As with any research project, there were roadblocks. At times, the model grouped similar players too closely, producing nearly identical predictions. To fix this, Price normalized the data using a technique called MinMaxScaler, which adjusted the range of values to better distinguish individual player characteristics. This helped the model generate more varied and accurate results. She also learned to scale up her computing needs, moving from OSC’s classroom Jupyter Notebooks to the more powerful Pitzer cluster when data loads grew too large.
Despite starting with zero machine learning experience, Price found the tools approachable.
“OSC made it so simple to get started. Our professor gave us Jupyter Notebooks to work with and everything was right there. It was all so well integrated and beginner-friendly,” said Price, who presented the results of her project at the spring 2025 OSC Research Symposium.
OSC’s resources make all the difference for the many students who enter the class with very little experience with data science, Hughes said.
“I use an intensive hands-on approach with Jupyter Notebook modules run through OSC’s Open OnDemand,” he said. “Early in the semester, I provide more structure and hints, but as the course progresses, students learn to solve problems on their own. It’s simple for beginners, yet close to what professional data scientists use every day. Students here are incredibly lucky to have access to a world-class computing platform in Ohio.”
What began as a playful spin on a class project grew into something much bigger. For Price, it was the first time she saw herself not just solving problems but building tools that could make sense of massive, complex data.
She is now planning to pursue a graduate degree in machine learning, with hopes of continuing to build models and personal projects.
“Before this class, I never considered data science as a career,” she said. “It turned out to be the most impactful course I’ve taken. Now I know this is the direction I want to go.”
Hughes said that’s exactly the kind of outcome he envisions for his course.
“I’ve worked with many students in both research and industry settings,” he said. “The skills students gain here can serve as a foundation for wherever they go next.”
Price’s story is evidence of the power of applied learning and accessible tools. With OSC’s resources and the flexibility of Physics 5680, she turned a personal interest into a technically rigorous project—which resulted in a clear new path for her.
Written by Lexi Biasi
The Ohio Supercomputer Center (OSC) addresses the rising computational demands of academic and industrial research communities by providing a robust shared infrastructure and proven expertise in advanced modeling, simulation and analysis. OSC empowers scientists with the services essential to making extraordinary discoveries and innovations, partners with businesses and industry to leverage computational science as a competitive force in the global knowledge economy and leads efforts to equip the workforce with the key technology skills required for 21st century jobs.
