A Multi-Task Probabilistic–Machine Learning Framework for Forensic Anthropological Identification: Integrating Osteology, Taphonomy, and Forensic Entomology

Abstract

Forensic identification commonly rests on estimation of biological sex, age at death, stature, and postmortem interval. This paper proposes a principled, end-to-end, multi-task probabilistic machine learning framework that fuses skeletal and dental measurements, radiographic descriptors, taphonomic variables, and entomological evidence to infer joint posterior distributions. The framework foregrounds calibrated probabilities and explicitly models uncertainty, offering a comprehensive approach to forensic anthropological identification.