PUMAA: Establishing a protocol for utilizing machine learning in forensic anthropological analyses.

Area of Science:

• Forensic Anthropology
• Machine Learning
• Computational Biology

Background:

• Machine learning (ML) applications are growing in forensic anthropology (FA).
• Current research lacks standardized protocols for ML model curation, utilization, and assessment.
• This gap hinders consistent and reliable application of ML in forensic analyses.

Purpose of the Study:

• To introduce PUMAA (A Protocol for Utilizing Machine Learning in Forensic Anthropological Analyses).
• To provide a standardized framework for forensic practitioners using ML models.
• To enhance the accessibility and understanding of ML concepts in FA.

Main Methods:

• Development of PUMAA, including a flowchart and checklist.
• Explanation of common supervised ML models in accessible terms with visuals.
• Evaluation of five key factors for assessing ML model performance.
• Discussion of reporting standards for seven ML model types.

Main Results:

• PUMAA offers a structured approach for ML model lifecycle management in FA.
• The protocol details essential factors for evaluating ML model performance.
• Accessible explanations and visual aids simplify complex ML concepts for practitioners.
• Strengths and limitations of various ML models are evaluated to guide selection and application.

Conclusions:

• PUMAA establishes an initial standard for ML implementation in forensic anthropology.
• The protocol aims to improve the rigor and reproducibility of ML-driven forensic research.
• Standardization through PUMAA will facilitate informed decision-making regarding ML model use in FA.