Predictors of early discharge in health facilities after birth in East African. a machine learning approach

Abstract

Code for the manuscript "Predictors of early discharge in health facilities after birth in East Africa: A machine learning approach" (under review at PLOS ONE). This repository contains Python code for data preprocessing, machine learning model training and tuning (LR, DT, RF, AdaBoost, XGBoost, KNN) using Optuna, SHAP interpretability, association rule mining (Apriori), and subgroup-stratified performance evaluation. Requires Python 3.10+ and libraries listed in requirements.txt. DHS data not included due to data use restrictions. See README.md for full details.