41
groups of horses using the information held within dental
architecture, specifically the enamel pattern of the occlusal surface of
the second premolar (P2) and third molar teeth (M3) of the maxillary
dental arcade. 2-D Geometric data was collected from photographs
of horse teeth using computer software (TPSdig2) to landmark eight
specific anatomical points on the tooth surface. Following
landmarking, Procrustes analysis was performed to separate size and
shape data.
The 2-D GMM data collection was analysed using advanced
statistical modelling techniques, including penalised logistic
regression, partial least squares regression, and linear discriminant
analysis, which attempt to identify the optimal linear combinations of
the variables for separating the samples into categories. As an
alternative to statistical modelling, this study used machine learning
tools such as support vector machines, naive Bayes, and random
forests, which can build complex non-linear predictors. Preliminary
statistical models and machine learning analysis of the occlusal
patterning of the P2 and M3 teeth in these samples showed some
population separation by site, while separation by domestication
status remained inconclusive. Overall, our results will determine
whether or not, even in combination with powerful statistical
modelling, the study of the two-dimensional shape of horse teeth via
GMM will be sufficient to definitively classify archaeological
samples according to their domestication status.
Достарыңызбен бөлісу: