Exploring the Relationship Between Cranial Non-Metric and Metric Traits For Ancestry Estimation
MetadataShow full metadata
Ancestry estimation is an important component in the discipline of forensic anthropology. Forensic anthropologists either visually assess skeletal remains through cranial macromorphoscopic traits or via craniometric analyses. Typically these two approaches are assessed separately as standalone methods. In this study, a variety of machine learning methods (decision tree analysis, random forest modeling, artificial neural networks, support vector machines, and linear discriminant functions) were applied to macromorphoscopic, craniometric, and combined (macro and metric) datasets to evaluate the classification accuracies of each and to explore how their individual and combined contributions may affect the estimation of ancestry. Overall, the random forest model performed the best out of the methods in two of the datasets with a classification accuracy of 95% for the metric data and 90% for the macromorphoscopic data. For the combined dataset, the support vector machines performed the best at 90%. The present study demonstrates the utility of these new methods contributing a greater wealth of information to group classification and also improving knowledge that these two data types can be combined into a single statistical analysis with classification accuracies of 90% and above for specific machine learning methods.