Genetic Programming for Age-at-death Estimation from the Pubic Symphysis

Enrique Bermejo, Antonio David Villegas, Javier Irurita, Sergio Damas, Oscar Cordón

Keywords: Genetic Programming, Interpretable Machine Learning, Age-at-deat estimation, Biological profile, Forensic Anthropology

TL;DR: This study proposes a semi-automatic method based on genetic programming using nine pubic symphysis traits to estimate age-at-death, achieving state-of-the-art performance and enabling new insights in forensic anthropology.

Abstract: Skeleton-based age-at-death estimation is an arduous task in human identification based on characteristics such as appearance, morphology or ossification patterns. This process is performed manually, although in recent years there have been several studies that attempt to automate it. This study proposes a semi-automatic method for estimating age-at-death using nine pubic symphysis traits derived from Todd's method. By employing genetic programming and symbolic regression, simple mathematical expressions are generated to estimate age-at-death. To address the imbalance in the data, oversampling methods are implemented. The method achieves state-of-the-art performance while maintaining interpretability, allowing validation of existing knowledge and discovery of new insights in forensic anthropology.

Submission Number: 6