Using machine learning on new feature sets extracted from 3D models of broken animal bones to classify fragments according to break agent

Abstract

Distinguishing agents of bone modification at paleoanthropological sites is at the root of much of the research directed at understanding early hominin exploitation of large animal resources and the effects those subsistence behaviors had on early hominin evolution. However, current methods, particularly in the area of fracture pattern analysis as a signal of marrow exploitation, have failed to overcome equifinality. Furthermore, researchers debate the replicability and validity of current and emerging methods for analyzing bone modifications. Here we present a new approach to fracture pattern analysis aimed at distinguishing bone fragments resulting from hominin bone breakage and those produced by carnivores. This new method uses 3D models of fragmentary bone to extract a much richer dataset that is more transparent and replicable than feature sets previously used in fracture pattern analysis. Supervised machine learning algorithms are properly used to classify bone fragments according to agent of breakage with average mean accuracy of 77% across tests.

Use and Misuse of Machine Learning in Anthropology

“…Since bones are one of the artifacts that occur in abundance at paleoanthropological sites, the identification of agents of bone breakage is essential to understanding how the site formed, how early humans evolved biologically and behaviorally, and how they interacted with their environment and with each other. However, longstanding debates over such identifications have yet to be resolved at important paleoanthropological sites such as are found in Dikika, Ethiopia (3.4 Ma) and Olduvai Gorge, Tanzania (1.8 Ma) [1], [2]. Some researchers have applied machine learning to feature sets that are traditionally used in taphonomic analysis and are based on qualitative features as observed by the analyst and measurements taken manually [3]- [7].…”

“…using simple flakes with straight cutting edges versus retouched flakes that have a more serrated cutting edge) [3], differentiating marks made on fleshed and defleshed bones [10], exploring how captivity and domestication of dog species affects the morphology of the traces they leave behind [11], testing the efficacy of different methodologies [12], and testing inter-and intra-observer variation during the process of feature extraction [4], [12]. Machine learning has also been applied to fracture patterns resulting from marrow extraction to identify whether carnivores or humans were responsible for breaking the bones [1], [6], [7].…”

“…Within the realm of fracture pattern analysis, machine learning has been applied to new feature sets resulting from recently developed methods for feature extraction from 3D models of bone fragments [1], [14]. These methods are highly accurate and replicable and can be applied to 3D models of any object.…”

“…In this section, we review recent applications of ML within our interdisciplinary consortium, the Anthropological and Mathematical Analysis of Archaeological and Zooarchaeological Evidence (AMAAZE) 1 . The motivating purpose of AMAAZE is to leverage advanced mathematical tools, particularly machine learning, time series analysis, and advanced geometric methods, to address fundamental questions in anthropology and human evolution that arise from the study of bones, lithics, and other artifacts, and thereby foster new and productive collaborations between mathematicians and anthropologists.…”

“…We have recently applied machine learning to a feature set extracted using the aforementioned tools [1]. The purpose of this study was to differentiate bone fragments broken by hominins using hammerstone and anvil from those broken by spotted hyenas through mastication.…”