Gender classification from anthropometric measurement by boosting decision tree: A novel machine learning approach

Area of Science:

• Anthropometry
• Machine Learning
• Biometrics

Background:

• Accurate gender classification is crucial in various fields, including forensics and personalized medicine.
• Anthropometric measurements offer a non-invasive method for human identification.
• Developing efficient algorithms for gender classification from body measurements is an ongoing research area.

Purpose of the Study:

• To develop and validate a gender classification algorithm using anthropometric data.
• To identify the most significant anthropometric variables for gender determination.
• To apply a boosting tree algorithm for high-accuracy gender classification.

Main Methods:

• Utilized a training dataset of twenty-five anthropometric measurements.
• Employed a boosting tree algorithm for classification.
• Applied a decision tree approach to generate classification rules and perform dimension reduction.

Main Results:

• Identified twelve significant anthropometric variables: chest diameter, waist girth, biacromial, wrist diameter, ankle diameter, forearm girth, thigh girth, chest depth, bicep girth, shoulder girth, elbow girth, and hip girth.
• Achieved a high accuracy rate of 98.42% for gender classification.
• Developed seven decision rule sets for dimension reduction.

Conclusions:

• The boosting tree algorithm effectively classifies gender using a reduced set of anthropometric measurements.
• Selected anthropometric variables are highly indicative of gender.
• The developed algorithm demonstrates significant potential for practical applications requiring gender identification.