Solving for Today Through Predictive Modeling Competition
How does a large company tackle complex business challenges in imaginative ways? One effective method is to tap into the creativity data-driven employees bring to the table, providing them development opportunities and generating outcomes to create the products that best serve our customers.
Bring on the friendly competition.
During the 2020 Virtual Predictive Modeling Competition, 75 companywide data scientists, data engineers and technology-oriented employees formed into 21 teams with one challenge: to develop a new model that helps our Claim organization more swiftly predict the severity of an accident at faster rates and improve payout predictions.
The eight-week competition gave participants the chance to work with new technologies and frameworks that deepened their skillsets and provided opportunities to collaborate across departments. The event was sponsored by Travelers Claim Business Intelligence and Analytics (CBIA) and Enterprise Data and Analytics (ED&A).
Teams relied on their knowledge of AWS cloud technology, graphics processing units, unstructured data sources and machine learning to analyze a range of information to develop and build their models.
According to the competition winners, reviewing data from customer interaction with claim handlers was paramount. “We didn’t reinvent the wheel, but our model predictions were sensible and could predict high severity claims with a high level of confidence,” said Murat Yasar, Analytics & Research, Business Insurance.
“I saw the competition as a great challenge. It was also a great way to network and share knowledge – which is really a precious opportunity, especially during the pandemic,” said Susan Ye, Data & Research, CBIA.
Travelers began these modeling competitions in 2016 with the goal of providing training opportunities around deep learning and image analytics. But the impact of the work goes far beyond training.
“The winning solution from the 2018 competition is still being used at Travelers – it’s like a Swiss army knife for data scientists,” said George Lee, who leads the Data Science team in ED&A’s AI Accelerator. “That’s our hope for this year’s winning solution.”