Attention: Due to COVID-19 restrictions, seat capacity at Prometric testing centers continues to be affected. If you are unable to find a seat for your exam, click here to see your options.

A Machine Learning Approach to Incorporating Industry Mortality Table Features into a Company's Insured Mortality Analysis

September 2019

The Product Development Section, the Modeling Section, the Reinsurance Section and the Financial Reporting Section announce the release of a new research report that introduces a novel framework for leveraging the “architecture” of an industry mortality table within a company’s predictive analytics-based insured mortality analysis. Authored by Marc Vincelli, the paper presents methodology for insurers to better model the relationships between mortality cells across ages and durations when faced with sparse experience data. One potential application of the approach is in the initial calibration of an industry mortality table, via its learned features, to a company’s own experience. 


A Machine Learning Approach to Incorporating Industry Mortality Table Features into a Company’s Insured Mortality Analysis

Appendix A Files

Appendix D Tables

Thank You

The Sponsors would like to thank the individuals who served on the Project Oversight Group:

June Quah – Chair

Mary Bahna-Nolan

Larry Bruning

Tom Farmer

Sara Goldberg

Andrew Harris

Michael Kula

Nora Li

Yuanjin Liu

Michael Niemerg

Yvonne Ren

Zachary Stenberg

Si Zhang

Mervin Kopinsky, SOA Experience Studies Actuary

Jan Schuh, SOA Sr. Research Administrator

Ronora Stryker, SOA Sr. Practice Research Actuary

Questions Or Comments?

If you have comments or questions, please send an email to

Disclaimer for Software

Important: The code, models, and files ("Software") is the property of the Society of Actuaries (SOA) and is protected under U.S. and international copyright laws.

The Software has been developed for the benefit of actuaries FOR EDUCATIONAL USE ONLY, although others may find it useful. SOA and the author make the Software available to individual users for their personal use on a non–exclusive basis. No commercial use, reproduction or distribution is permitted whatsoever.

SOA, and the author make no warranty, guarantee, or representation, either expressed or implied, regarding the Software, including its quality, accuracy, reliability, or suitability, and HEREBY DISCLAIM ANY WARRANTY REGARDING THE SOFTWARE'S MERCHANTABILITY OR FITNESS FOR ANY PARTICULAR PURPOSE.  SOA, and the author make no warranty that the Software is free from errors, defects, worms, viruses or other elements or codes that manifest contaminating or destructive properties. In no event shall the SOA or the author be liable for any damages (including any lost profits, lost savings, or direct, indirect, incidental, consequential or other damages) in connection with or resulting from the use, misuse, reliance on, or performance of any aspect of the Software including any instructions or documentation accompanying the Software. The  SOA, and the author make no representation or warranty of non–infringement of proprietary rights of others with respect to the Software. The entire risk as to the uses, outputs, analyses, results and performance of the Software is assumed by the user. This Disclaimer applies regardless of whether the Software is used alone or with other software.

The models, any accompanying code, documentation, and methodologies contained herein do not represent an official position, statement, or endorsement on behalf of the Society of Actuaries or its members, nor should the material be construed to do so. It is the product of a research effort commissioned by the Society of Actuaries. The models and code are neither intended to preclude the use of other methodologies for any purpose nor provide a statement or position on the use, application, or preferability of other methodologies as compared to the methodology described herein.