Osteoporosis Feature Selection and Risk Prediction Model by Machine Learning Using a Cross-Sectional Database.

Area of Science:

• Biomedical Informatics
• Data Science in Healthcare
• Osteoporosis Research

Background:

• Osteoporosis poses a significant public health challenge.
• Accurate risk factor identification is crucial for early detection and prevention.
• Existing models may not fully capture sex-specific risk differences.

Purpose of the Study:

• To validate machine learning (ML) for osteoporosis risk factor selection.
• To identify sex-specific differences in ML-driven feature selection for osteoporosis.
• To develop accurate ML-based predictive models for osteoporosis.

Main Methods:

• Utilized data from 3,484 participants in the Korea National Health and Nutrition Examination Survey.
• Applied logistic regression, random forest, gradient boosting, adaptive boosting, and support vector machine for feature selection.
• Analyzed 968 observed features to identify preliminary risk factors for osteoporosis.

Main Results:

• Body mass index, alcohol consumption, and dietary surveys were common risk factors for both sexes.
• Age, smoking, and blood glucose levels showed sex-specific differences in ML-based feature selection.
• Receiver Operating Characteristic (ROC) analysis indicated no significant difference in model performance between genders.

Conclusions:

• Machine learning successfully identified osteoporosis risk factors, accounting for sex-based variations.
• Data preprocessing and feature selection are critical for enhancing ML model accuracy in osteoporosis prediction.
• The study underscores the importance of considering sex-specific factors in osteoporosis risk assessment.