EXPRESS: Age estimation of bloodstains based on Kernel Extreme Learning Machine algorithm and Near-infrared...

Area of Science:

• Forensic Science
• Analytical Chemistry
• Machine Learning

Background:

• Bloodstains are critical forensic evidence, but traditional age determination methods are destructive and costly.
• Accurate bloodstain age estimation is vital for crime scene investigations.
• There is a need for rapid, non-destructive, and accurate bloodstain age determination techniques.

Purpose of the Study:

• To develop a rapid, non-destructive, and accurate method for estimating bloodstain age.
• To integrate spectral data with machine learning for predictive modeling of blood aging.
• To compare the performance of the Kernel Extreme Learning Machine (KELM) algorithm with other models.

Main Methods:

• Bloodstains were applied to various substrates to minimize background interference.
• Near-infrared (NIR) spectral data were extracted non-destructively from bloodstains.
• Predictive models were developed using machine learning, including KELM, Partial Least Squares Regression (PLS-R), Random Forest (RF), and Back-propagation neural network (BP).

Main Results:

• The KELM model demonstrated superior performance with a determination coefficient (R²) of 0.94 on the test set.
• KELM achieved a lower root mean square error of prediction (RMSEP) of 2.54 days compared to other models.
• The KELM algorithm yielded the minimum mean absolute percentage error (MAPE) of 2.45% on the validation set.

Conclusions:

• The proposed method using NIR spectral data and KELM accurately estimates bloodstain age.
• This non-destructive technique offers a significant advancement over traditional methods.
• The study provides a new technological reference for forensic bloodstain analysis.