bigmike2307/maternal-health-fairness-africa

# Maternal Health Fairness Analysis Across African Countries Investigating algorithmic fairness in maternal health risk prediction models using Demographic and Health Survey (DHS) data from multiple African countries. ## Overview This project examines how machine learning models for maternal health risk prediction perform across different demographic subgroups in African populations. We evaluate fairness metrics across: - **Wealth quintiles** (poorest to richest) - **Urban/rural residence** - **Educational attainment levels** - **Geographic regions within countries** ### Research Questions 1. Do maternal health risk prediction models perform equally well across wealth quintiles? 2. Are there systematic disparities in model accuracy between urban and rural populations? 3. How does model performance vary across different African countries? 4. What fairness interventions can reduce identified disparities? ## Countries Analyzed | Country | DHS Survey Year | Sample Size | |---------|-----------------|-------------| | Nigeria | 2018 | ~40,000 women | | Kenya | 2022 | ~15,000 women | | Ghana | 2022 | ~9,000 women | | Uganda | 2016 | ~18,000 women | | Tanzania | 2022 | ~13,000 women | ## Project Structure ``` maternal-health-fairness-africa/ ├── data/ # DHS datasets (not tracked in git) │ ├── raw/ # Original DHS files │ └── processed/ # Cleaned datasets ├── notebooks/ │ ├── 01_data_exploration.ipynb │ ├── 02_feature_engineering.ipynb │ ├── 03_model_training.ipynb │ ├── 04_fairness_evaluation.ipynb │ └── 05_cross_country_comparison.ipynb ├── src/ │ ├── data_loader.py # DHS data loading utilities │ ├── preprocessing.py # Data cleaning functions │ ├── features.py # Feature engineering │ ├── models.py # ML model definitions │ └── fairness_metrics.py # Fairness evaluation functions ├── results/ │ ├── figures/ # Visualization outputs │ └── tables/ …