Professional Insight

Travelers Actuaries Cook Up Some Predictive Modeling Fun

Panelists from “The Travelers Predictive Modeling Cooking Show” are (left to right) Melissa Schenck, ACAS; Megan Camanocha, ACAS; Nathan Hubbell, FCAS; Chad Wilson, FCAS; and Richard Sutherland, FCAS.

Predictive modeling and cooking have a lot in common. There is choosing quality ingredients, apportioning them just right and applying the relevant tools and techniques to create rewarding results.

Just as cooking shows have transformed the culinarily challenged into foodies, actuaries from Travelers are applying the same approach to encourage predictive modeling. The aptly named session, “The Travelers Predictive Modeling Cooking Show,” is receiving rave reviews.

Here’s what attendees at various CAS meetings are saying: “Really informative!” “My favorite session of the day.” “Technically sound presentation done in a casual way.”

The show, which originally began as an internal training class, features an interactive competition between two teams, “Slice and Dice” and “Occam’s Razor,” which face off to develop the best model.

Chad Wilson, an inland marine actuary, envisioned the idea of the show after he sat through a 20-hour online predictive modeling course. Desiring to present the principles, techniques and process of predictive modeling in a quicker and more engaging manner, he began considering what a predictive modeling session in a cooking show format would be like.

With the enthusiasm and support from other Travelers actuaries, his idea began taking root. “We developed the event to help the audience members understand the basics of modeling and understand the modeling process,” says Melissa Schenck, a customer analytics actuary. “We want the audience to understand the handoff in the process to create a better predictive model,” she explains.

Other team members who helped develop or participated in the session include modeling actuaries Megan Camanocha, Nathan Hubbell, Laura Johnson and Jon Sanders; Rick Sutherland, a product liability actuary; Sovanna Ly, a workers’ compensation actuary; and Priyanka Shah, a business intelligence consultant. Sam Srini, director of advanced analytics data engineering, and Cheryl Philstrom, senior administrative assistant, are also part of the team.

The show begins when Wilson takes the stage as the master of ceremonies, explains the competition and introduces the three to seven actuaries available to participate. The chef and sous chef demonstrate how to prep for predictive modeling through exploratory analysis to find underlying relationships in the data.

Members of the audience each get a piece of the data and play along, submitting their own predictions to build a collective, group-sourced, audience model.

As both teams undergo the modeling process, Wilson asks why a data set, a.k.a. the ingredients, is being imported in the model and the reasoning behind selecting certain variables. “We get into feature engineering,” Wilson describes, explaining, “[There is a] ‘feature of the day’ from creating new data from the data you have.” The session also includes lift charts and confusion matrices.

“The most important piece is the drama surrounding the data,” Schenck says. The dish is “taste-tested,” Wilson adds, when the models are checked according to standard practices. “In business, the modeler will tell you the reasons why it works and try to explain the measurable metrics for how effective it is,” he notes.

“The Travelers Predictive Modeling Cooking Show” has been on tour offering seven shows in five cities over the past year at conferences for underwriters, actuaries and regulators. Its next stop will be at the Gamma Iota Sigma International Conference in Dallas on Friday, September 27.


Annmarie Geddes Baribeau is a frequent AR contributor. Her blog can be found at www.insurancecommunicators.com.