Comparing Machine Learning Classifiers and Linear/Logistic Regression to Explore the Relationship between Hand...

Area of Science:

• Biometrics and Forensics
• Human Physiology
• Machine Learning Applications

Background:

• Understanding physiological measurements in relation to demographic data is crucial for biometrics and forensics.
• The human hand offers a rich source of physiological data with potential demographic correlations.
• Existing methods may not fully leverage the predictive power of hand measurements for demographic estimation.

Purpose of the Study:

• To explore the relationship between human hand measurements and demographic features.
• To assess the predictive accuracy of linear regression and machine learning classifiers for demographic estimation from hand data.
• To identify key hand features that underpin demographic relationships for practical applications.

Main Methods:

• Collected physiological measurements from human subjects, focusing on hand dimensions.
• Applied linear regression models to analyze the relationship between hand features and demographics.
• Utilized machine learning classification algorithms to predict demographic variables (sex, height, weight, foot size) from hand measurements.

Main Results:

• Machine learning classifiers significantly outperformed linear regression in predicting sex, height, weight, and foot size.
• Accurate demographic predictions were achieved across various data-range bin sizes.
• Key hand features driving these demographic relationships were identified and validated.

Conclusions:

• Human hand measurements are strong predictors of demographic characteristics like sex, height, and weight.
• Machine learning techniques offer superior performance over linear regression for demographic prediction using hand biometrics.
• The identified hand-feature relationships have broad applicability in biometric and forensic identification systems.