Transfer learning for mortality risk: A case study on the United Kingdom.

This study introduces a transfer learning framework to address data scarcity in mortality risk prediction for the UK, where local mortality data is unavailable. By leveraging a pretrained model built from data across eight countries (excluding the UK) and incorporating synthetic data from the countr...

Full description

Saved in:
Bibliographic Details
Main Authors: Asmik Nalmpatian, Christian Heumann, Levent Alkaya, William Jackson
Format: Article
Language:English
Published: Public Library of Science (PLoS) 2025-01-01
Series:PLoS ONE
Online Access:https://doi.org/10.1371/journal.pone.0313378
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This study introduces a transfer learning framework to address data scarcity in mortality risk prediction for the UK, where local mortality data is unavailable. By leveraging a pretrained model built from data across eight countries (excluding the UK) and incorporating synthetic data from the countries most similar to the UK, our approach extends beyond national boundaries. This framework reduces reliance on local datasets while maintaining strong predictive performance. We evaluate the model using the Continuous Mortality Investigation (CMI) dataset and a Drift model to address discrepancies arising from local demographic differences. Our research bridges machine learning and actuarial science, enhancing mortality risk prediction and pricing strategies, particularly in data-poor settings.
ISSN:1932-6203