‘Spark Penn State’ Takes Top Prize At 48-Hour DataFest
Spending 48 hours working on statistical problems and data analysis isn’t the typical weekend for a college student, but it’s a dream weekend for students like those in Spark Penn State.
The Penn State DataFest’s inaugural competition was held this weekend, pitting student teams together in hopes of solving some statistical problems. The teams were presented with data from Edmunds.com, a car research site for automobile consumers.
“It was very open-ended,” said Dom Mirabile, a member of the winning team. “We were tasked with finding novel insights and working on making the user experience easier. Edmunds gave us an overview of their business model and the data they were getting.”
Mirabile is a member of Spark Penn State, an organization focused on statistical analytics and data. The organization is brand new, founded within the last year and made up of just seven members. Spark didn’t let its infancy hold it back, winning the overall award and the creativity award at DataFest. The team primarily worked with targeted advertising, trying to utilize data to provide web ads targeted directly to users based on their behavior on the website.
“We chose to apply machine learning around their advertising, which is where they make the majority of their revenue,” Mirabile said. “We looked into the feasibility of a data-driven targeted ad model. We worked on a system that evaluates how you interact with the site to determine what ads you see.”
That might be difficult to understand, but the general concept is to establish a system that can educate itself to better target advertising.
“The best example is a spam filter,” Mirabile said. “It uses a data set to learn how to classify something. For instance, the more you identify spam the better it gets at recognizing it without you having to tell it something is spam.”
Spark took that concept and applied it directly to Edmunds. As users navigate the site, type in search terms, and view specific cars, data is collected that paints a picture of who the user is. That data can then learn what advertising works best for what type of user, and it becomes better as time goes on and the statistical pool grows.
After reflecting on the victory, Mirabile was excited that his five-person team took the top prize at DataFest after 48 hours of hard work.
“We were excited. It says a lot about how creativity is important when thinking about a complex problem,” he said. “The whole thing was really well planned. There were mentors that were extremely helpful and a lot of good food all weekend.
Spark is less than a year into its work on campus, and Mirabile feels that the organization’s success at DataFest should encourage other students to get involved going forward.
“I feel like a lot of people are hesitant to do something like this because they don’t have the skills,” he said, “but we prove that shouldn’t hold you back.”