University at Buffalo (UB)
Making sense of 1.3 billion eBay listings
UB students participate in eBay Machine Learning Challenge
From left, graduate students Souvik Das, Sougata Saha and Rabiraj Banerjee are participating in the first-ever eBay Machine Learning Challenge, to help the industry giant find ways to make sense of its massive amounts of unstructured data.
by Nicole Capozziello
Published December 12, 2019
Since launching nearly 25 years ago, online marketplace eBay has grown astronomically, currently hosting over 1.3 billion listings of everything from model cars to Cars DVDs to actual cars.
Sorting all of these listings is a vast and ever-changing undertaking, one that eBay has historically relied on in-house machine learning professionals to manage.
This year, however, the company decided to host an exciting competition: the first-ever eBay Machine Learning Challenge. Student teams from four elite schools – UB, NYU, Stanford and the University of Texas at Austin – were chosen to participate in the challenge, which is to create a product catalog for eBay.
When eBay put out the call for the challenge in early fall, students from the Department of Computer Science and Engineering were interested. Graduate students Sougata Saha, Souvik Das and Rabiraj Banerjee reached out to eBay and began the challenge on October 11. The team will work on the challenge until March, with the winning team being announced on March 25, 2020.
“I am excited about the real-world challenges that this competition exposes the students to,” says Varun Chandola, assistant professor in the Department of Computer Science and Engineering, and mentor for the team. “Working with real data and practical constraints allows the students to deal with different types of challenges that they might not have encountered in classroom education and activities.”
“It’s a great opportunity to dive deeper into the field of machine learning and deep learning,” says Saha. “Also, given that the competition is four months long, we’ll have enough time to research and try out different approaches.”
The team began by doing exploratory analysis on the dataset. Currently they are reading up on some deep learning literature, gaining an understanding of individual problem components. Over winter break, they intend to start their experiments. “We’ve identified a couple of potential approaches to tackle the problem,” says Saha. “It’ll definitely keep us busy during the long winter break.”
Students on the winning team will be offered a paid 12-week internship at eBay’s San Jose headquarters for the summer of 2020. The internship program would give students the benefits of working at a leader in tech, as well as provide them with the unique opportunity to apply their Machine Learning models in a real-life setting.
In addition to the collaborative project providing eBay new ideas for improving its platform by giving rising technologists access to this type of data, the company also hopes the competition will “pique academic curiosity within machine learning” spurring, “more research in the ecommerce domain powered by a real-world ecommerce dataset.”