Travelers Predictive Modeling Competitions: Bridging Business Trends with Learning Experiences
The Travelers Enterprise Data & Analytics (ED&A) team recently held its ninth-annual Predictive Modeling Competition with record participation, yielding a winning model poised to enhance the understanding of auto bodily injury trends over time.
Hosted by Business Insurance (BI), the competition challenged teams to develop predictive models using internal policy-level and claim-level data. These models were aimed at providing accurate predictions and explanations to elucidate the drivers of trends within the business.
Kevin Zech, Senior Director and Data Scientist for BI, emphasized that the Modeling Competition isn’t just for data scientists.
“It’s a great learning experience for predictive modeling concepts, but also an excellent way to keep on top of the leading tools and technology in the industry,” he said.
Participants were given structured access to necessary tools and technologies, including gitlab repositories, clean datasets and cloud resources, enabling them to focus on the core task of building and refining models.
Claiming the top spot was team “import pandas as pd,” a nod to programming language Python. Consisting of Hansook Choi and Ben Veasey from ED&A, along with Will Duke from Claim, the team’s model identified key words and patterns indicative of complex and often costly claims.
Ben Veasey, Senior Associate Data Scientist with ED&A, highlighted the team’s approach of simplifying their model for easier optimization.
“We focused on delving into the data and crafting features that could better understand and rectify the data, leading to improved prediction of claim payouts,” he explained.
Despite being new to analyzing Claim Notes data, the team leveraged their experience in processing text data and building machine learning models to develop an automated, quantitative method for identifying claim issues.
Elvis Martinez, Director & Data Scientist for Personal Insurance, celebrated winning third place alongside fellow Take it EC teammate High Seng Chai. “The real win for me is the opportunity to collaborate outside my usual scope,” Elvis remarked. “Working with individuals like Chai allows me to learn from diverse ideas and approaches.”
The Predictive Modeling Competition is one of many initiatives hosted by Travelers to foster collaboration and innovation among employees to solve real-world business challenges. Over the years, outcomes from such events have influenced various products and procedures benefiting customers, employees and the insurance industry at large.
