Smarter Segmentation for the Lazy Actuary

Segmentation in reserving is often informed and limited by time pressures, resource constraints, and “classic” definitions of segmentation. By using advanced modeling techniques, actuaries can identify different cohorts of claims – those most likely to develop poorly – and reserve for them separately. This session will look at how machine learning approaches can allow a reserving team to improve their reserve estimates, not by segmenting more, but by segmenting smarter.

  • Date:Monday, September 13
  • Time:5:00 PM - 6:15 PM EDT
  • Session Type:Concurrent Session
  • Session Code:AR-3
  • Learning Objective 1:Construct GBMs on loss data for the purpose of claims segmentation.
  • Learning Objective 2:Incorporate geographic dimensions into the GBM claims analysis.
  • Learning Objective 3:Use the GBM results for specifying claims cohorts (including interpretation).
  • Learning Objective 4:Demonstrate advantages and impact of improved segmentation.
  • Level of Knowledge:Level 2: General knowledge of the subject
  • Moderator:Shawn Balthazar
Kacie Kiel, ACAS
CNA Insurance Companies
Serhat Guven, FCAS
Willis Towers Watson