| TITLE | A Contrast-Tree-Based Approach to Two-Population Models |
| JOURNAL | Risk - Life Insurance and Pensions: Latest Advances and Prospects |
| SUBJECT | Actuarial Sciences |
| DATE | September 2024 |
| AUTHOR | Matteo Lizzi |
ABSTRACT
This paper investigates the use of a machine learning methodology, called Contrast Trees, to produce life tables for small populations. Specifically, in the context of two-population models, a modification to the Estimation Contrast Boosting technique is introduced to derive projected life tables for small populations, based on national projected life tables. The methodology was tested using data from the Human Mortality Database, with the Italian population used as the reference population, and the Austrian, Slovenian, and Lithuanian populations as the small populations. The results show greater prediction accuracy compared to rescaling of mortality tables, which is the methodology traditionally used in Italian actuarial practice.
ACCESS FULL ARTICLE
ACCESS FULL ARTICLE
Matteo Lizzi - Core Faculty
Matteo Lizzi is a member of the Core Faculty at MIB Trieste School of Management, in the Actuarial Sciences area. He is a researcher affiliated with the Center for Insurance Research "Ermanno Pitacco".
Lizzi is a fully qualified member of O.N.A. – the Italian National Order of Actuaries. With eight years of experience in the insurance industry, primarily as a life actuary, he also holds a Ph.D. in Actuarial Sciences from Sapienza University of Rome, where his thesis focused on "Improving Mortality Diagnostics and Estimation through Contrast Trees."
His research interests center on applying machine learning techniques to actuarial sciences.
