Integrating machine learning and statistical analysis to forecast insufficient physical activity trends using...

Area of Science:

• Global Health
• Epidemiology
• Public Health Policy

Background:

• Insufficient physical activity represents a significant global health challenge.
• Existing data reveal disparities in physical activity levels across different sexes and geographical regions.
• Understanding these inequalities is critical for developing effective public health strategies.

Purpose of the Study:

• To examine global inequalities in insufficient physical activity based on sex and region.
• To project future trends in physical activity levels.
• To identify regions and demographic groups most affected by insufficient physical activity.

Main Methods:

• Analysis of World Health Organization data using a Linear Mixed-Effects Model.
• Assessment of the impact of sex, time, and region on physical activity levels.
• Forecasting future trends using machine learning models including XGBoost, Random Forest, Multi-Layer Perceptron, and Support Vector Regression.

Main Results:

• Females exhibit higher and more variable rates of insufficient physical activity compared to males.
• Prevalence is lower in Africa and South-East Asia, but higher in the Americas and Eastern Mediterranean.
• Physical activity trends have worsened over time, especially for females and in high-prevalence regions.

Conclusions:

• Significant disparities in physical activity persist across sex and geographical regions.
• Machine learning model outputs vary, indicating uncertainty in future projections and the need for ensemble approaches.
• Targeted, gender-responsive public health interventions are essential to combat insufficient physical activity.