Principal component and linear discriminant analyses for the classification of hominoid primate specimens based on...

Area of Science:

• Paleoanthropology
• Primate Morphology
• Machine Learning Applications

Background:

• Understanding primate evolution and fossil hominoid relationships is crucial.
• Triquetrum bone shape offers potential for distinguishing primate groups.
• Previous methods lacked quantitative accuracy for comparative morphology.

Purpose of the Study:

• To test machine learning for classifying extant primates based on triquetrum shape.
• To apply this classification tool to infer affinities between extant primates and fossil hominoids.
• To assess the accuracy of principal component analysis (PCA) and linear discriminant analysis (LDA) in this classification task.

Main Methods:

• Utilized unsupervised (PCA) and supervised (LDA) machine learning pipelines.
• Extracted shape data from triquetrum bones of extant primates.
• Classified extant primates and subsequently applied the model to fossil hominoid specimens.
• Employed PCA and uniform manifold approximation and projection (UMAP) for data visualization.

Main Results:

• Achieved a high F1-score of 0.90 in classifying extant primate specimens using both PCA and LDA.
• The machine learning model accurately reflected known taxonomy and locomotor behaviors of fossil hominoids.
• Demonstrated the effectiveness of triquetrum shape analysis for distinguishing primate groups.

Conclusions:

• Machine learning applied to triquetrum shape data provides a highly accurate classification tool for extant primates.
• This method quantitatively compares fossil and extant primate morphology, supporting taxonomic differentiation.
• The approach aids in the interpretation of fossil remains and understanding evolutionary affinities.