Functional data geometric morphometrics with machine learning for craniodental shape classification in shrews.

Area of Science:

• * Zoology and Evolutionary Biology
• * Morphometrics and Biostatistics

Background:

• * Accurate species classification is crucial for understanding biodiversity and evolutionary relationships.
• * Traditional morphometric methods may not fully capture complex shape variations.

Purpose of the Study:

• * To introduce and evaluate a functional data analysis approach for morphometrics (FDGM) in shrew species classification.
• * To compare the efficacy of FDGM against classical geometric morphometrics (GM).
• * To identify the optimal craniodental view for species discrimination.

Main Methods:

• * Collected 2D landmark data from 89 shrew crania across dorsal, jaw, and lateral views.
• * Applied FDGM by converting landmark data into continuous curves represented by basis functions.
• * Utilized Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) for classification, comparing FDGM and GM outputs.
• * Evaluated four machine learning algorithms (naïve Bayes, SVM, random forest, GLM) with both methods.

Main Results:

• * FDGM demonstrated superior performance in classifying the three shrew species compared to GM.
• * The dorsal craniodental view was identified as the most effective for distinguishing between S. murinus, C. monticola, and C. malayana.
• * Machine learning models showed better predictive accuracy when using FDGM-derived scores.

Conclusions:

• * FDGM provides a powerful and more sensitive tool for morphometric analysis and species classification.
• * The dorsal view of the cranium is a key morphological region for differentiating these shrew species.
• * This approach has significant implications for taxonomic research and conservation efforts in the region.