Objectifying evidence evaluation for gunshot residue comparisons using machine learning on criminal case data.

Area of Science:

• Forensic Science
• Analytical Chemistry
• Machine Learning

Background:

• Comparative gunshot residue (GSR) analysis is crucial for forensic investigations, aiming to link GSR particles from a suspect to a specific firearm or scene.
• Current GSR analysis relies on expert interpretation of elemental composition, which can be subjective.
• There is a need for more objective and empirically grounded methods to support forensic experts in GSR interpretation.

Purpose of the Study:

• To develop and validate a likelihood-ratio (LR) system for comparative gunshot residue analysis.
• To enhance the objectivity and empirical grounding of GSR interpretation in forensic casework.
• To leverage machine learning models for analyzing high-dimensional GSR data.

Main Methods:

• Utilized statistical models from machine learning literature to construct the LR system.
• Employed a calibration step to ensure well-calibrated LR outputs.
• Developed and validated the system on both casework data and an independent dataset of cartridge data.

Main Results:

• The developed LR system demonstrated good performance on both casework and independent cartridge datasets.
• The system provides a more empirically grounded interpretation of electron microscopy analysis results for GSR.
• The findings support the potential of machine learning in forensic science for objective evidence evaluation.

Conclusions:

• The constructed LR system shows promise for supporting forensic experts in gunshot residue analysis.
• Further research and validation are necessary before widespread implementation in casework.
• Machine learning offers a powerful approach to enhance the interpretation of complex forensic data.