An unsupervised probabilistic net for health inequalities analysis

Area of Science:

• Public Health
• Data Science
• Computational Statistics

Background:

• Identifying health inequalities is crucial for global health promotion.
• Existing methods may not fully capture complex health status variations across countries.
• Understanding health disparities requires robust analytical tools.

Purpose of the Study:

• To introduce an unsupervised probabilistic net (UPN) for identifying health inequalities.
• To categorize countries based on health indicators and quantify inter- and intra-cluster inequalities.
• To utilize virtual objects for uncovering hidden patterns and informing health promotion strategies.

Main Methods:

• Employing an unsupervised probabilistic net (UPN) to estimate probability density functions of health indicators.
• Clustering countries with similar health statuses.
• Calculating intercluster inequalities using Mahalanobis distance and intracluster inequalities via cluster diversity.
• Introducing the concept of virtual objects to represent typical health statuses.

Main Results:

• The UPN successfully categorized countries into distinct health status clusters.
• Mahalanobis distance effectively identified intercluster health inequalities.
• Cluster diversity quantified intracluster health inequalities.
• Virtual objects were extracted, representing typical health profiles and hidden data knowledge.

Conclusions:

• The UPN provides a novel approach to identifying and quantifying health inequalities.
• Virtual objects offer valuable insights for social scientists and health promotion planning.
• This methodology can help establish realistic health promotion targets for countries facing health challenges.