This spring, LeBow’s master of science in Business Analytics program hosted its first Datathon, a challenge for student teams to complete analytics projects using open-source datasets. Launched as the “Dragon Datathon,” the initial experience was presented as an online challenge but will be offered in-person for future events, ensuring high-energy collaboration and a 24-hour turnaround of solutions.
The task: identify a business problem requiring analysis of data obtained in an open-source fashion, turn the business problem into a series of statistical questions, design a technical approach and create solutions to solve the statistical problem at hand.
Assessed by faculty on the scope of their analysis, implementation plan, solutions and expected impact, the winning team members were Koba Khitalishvili, Vivek Ghelani and Shantanu Saha. The group chose a competition posted by Santander Bank on Kaggle, a platform for analytics and predictive modeling competitions. The students concentrated on classifying banking customers as satisfied or unsatisfied based on 370 given features – a number the students were able to reduce to 36 after focusing on dimensionality reduction and analysis of the data. They were also able to increase the accuracy of their classification model.
“Competitions like these are great way to understand machine learning and open source tools,” said Khitalishvili. “We were able to get hands-on experience of data mining and classification.”