Exploring Ensemble Learning Techniques for Infant Mortality Prediction: A Technical Analysis of XGBoost Stacking...

Area of Science:

• Public Health
• Data Science
• Machine Learning

Background:

• Infant mortality is a critical public health concern.
• Accurate prediction of infant mortality is vital for targeted interventions.
• Factors like maternal education, prenatal care, nutrition, and environment influence infant mortality.

Purpose of the Study:

• To predict infant mortality rates in India.
• To evaluate the effectiveness of ensemble learning techniques for this prediction.

Main Methods:

• Ensemble learning techniques including eXtreme Gradient Boosting (XGBoost), stacking, adaptive boosting, and bagging were employed.
• Data from national surveys and demographic studies in India were preprocessed.
• Models were optimized using k-fold cross-validation and performance metrics were analyzed.

Main Results:

• XGBoost demonstrated superior performance with 98.75% accuracy, 98.56% precision, and 98.24% recall.
• Ensemble models effectively predicted infant mortality, with XGBoost excelling in handling complex data relationships.
• Significant regional disparities in infant mortality predictions were observed across India.

Conclusions:

• Ensemble learning models, especially XGBoost, are highly effective for predicting infant mortality in India.
• Targeted interventions are necessary in high-mortality regions, such as the Northeast.
• Improving maternal education, prenatal care access, and reducing socioeconomic disparities are critical.