Comparing Machine Learning Models for Neonatal Mortality Prediction: Insights from a Modeling Competition

This study explores the performance of various machine learning techniques in predicting neonatal mortality in NICUs. Through a modeling competition, five teams employed methods such as logistic regression, neural networks, random forest, CatBoost, and XGBoost on a shared dataset of over 6,000 NICU admissions. Surprisingly, simpler models like logistic regression outperformed more complex ones, highlighting the importance of method selection based on data characteristics and interpretability. The findings emphasize the value of explainable AI for clinical applications and provide a roadmap for advancing neonatal research using ML.

PUBLICATIONS