Studying infant mortality: A demographic analysis based on data mining models

Area of Science:

• Public Health
• Data Science in Healthcare
• Demographics

Background:

• Child mortality under five years remains a significant global health challenge, particularly in developing nations like Pakistan and Ethiopia.
• Despite global efforts, high mortality rates persist, necessitating advanced analytical approaches for effective intervention.
• Predictive analytics offers a powerful tool for understanding and mitigating child mortality trends.

Purpose of the Study:

• To identify and categorize the critical factors contributing to child mortality in Pakistan and Ethiopia using data mining techniques.
• To develop a predictive model for child death rates based on demographic and health survey data.
• To highlight the importance of data-driven insights for improving infant health outcomes.

Main Methods:

• Utilized datasets from the Pakistan Demographic Health Survey and Ethiopian Demographic Health Survey.
• Applied various data mining techniques including Bayesian network, J-48 (tree), PART (rule induction), random forest, and multi-level perceptron.
• Evaluated the performance of multiple classifiers to determine the most accurate predictive model for child mortality.

Main Results:

• Analysis of 12,654 (Pakistan) and 12,869 (Ethiopia) records identified key influencing factors on child mortality.
• The best performing model achieved an average accuracy of 97.8% in forecasting child death frequency.
• The developed model demonstrates the capability to estimate under-five mortality rates in the studied regions.

Conclusions:

• Data mining techniques effectively identify critical factors driving child mortality in Pakistan and Ethiopia.
• A highly accurate predictive model for child mortality has been developed, offering valuable insights for public health interventions.
• An online forecasting tool based on this research is recommended to aid healthcare strategies and reduce child deaths.