Combining machine learning algorithms and geometric morphometrics: A study of carnivore tooth marks

Abstract

Since the 1980s an intense scientific debate has revolved around the hunting capacities of early hominin populations and the behavioral patterns of carnivores sharing the same ecosystem, and thus competing for the same resources. This debate, commonly known as the hunter-scavenger debate, fostered the emergence of a new research line into the Bone Surface Modifications (BSMs) produced by both taphonomic agents. Throughout the following 20 years, multiple studies concerning the action of carnivores have been developed, with a particular focus on the oldest archaeological sites in East Africa. Recent technological advances applied to taphonomy have provided new insight into carnivore BSMs. A newly developed part of this work relies on Geometric Morphometrics (GMM) studies aimed at discerning carnivore agency through the morphologic characterization of tooth scores and pits. GMM studies have produced promising results, however methodological limitations are still present. This paper presents the first combined application of Machine Learning (ML) algorithms and GMM to the analysis of carnivore tooth marks, generating classification rates of 100% between carnivore species in some cases.