Machine learning-based classification of population affinity in two North Indian populations using morphological...

Area of Science:

• Forensic Science
• Anthropology
• Computer Science

Background:

• Forensic odontology utilizes dental evidence for personal identification.
• Morphological Dental Traits (MDTs) aid in classifying population groups for forensic identification.
• Machine Learning (ML) offers advanced analytical capabilities for forensic applications.

Purpose of the Study:

• To classify population affinity between two North Indian groups (Khasas and Kolis) using MDTs.
• To evaluate and compare the performance of various ML models for this classification task.
• To identify the optimal ML model for population affinity estimation based on dental features.

Main Methods:

• Collected dental casts from 434 participants (18-40 years) from two distinct North Indian populations.
• Observed and recorded Morphological Dental Traits (MDTs).
• Employed Recursive Feature Elimination (RFE) for feature selection and trained ML models including SVM, LR, DT, RF, and GBM.

Main Results:

• Logistic Regression (LR) achieved the highest classification accuracy at 75.8%, followed by Support Vector Machine (SVM).
• Decision Tree (DT) and Gradient Boosting (GBM) models showed the lowest accuracy at 64.3%.
• ROC-AUC and F1 score analyses corroborated LR as the top-performing model.

Conclusions:

• Machine learning approaches, particularly Logistic Regression, are effective for population affinity estimation using dental features in forensic odontology.
• This method can significantly support disaster victim identification and crime scene investigations involving dental remains.
• The study highlights the utility of ML in enhancing the precision of forensic identification processes.