nnettee87/Forecasting-Nutrition-Crises-in-South-Sudan

# Forecasting-Nutrition-Crises-in-South-Sudan Machine learning models for predicting nutrition crises in South Sudan using climate, conflict, and market data. ## 📋 Project Overview This project implements machine learning models to forecast nutrition crises (GAM ≥15%) in South Sudan by integrating climate, conflict, and market data. The research demonstrates both the potential and limitations of data-driven approaches for humanitarian early warning systems. ## 🎯 Research Question "Can machine learning models using climate, conflict, and market data reliably predict nutrition crises in South Sudan counties?" ## 📊 Key Findings ### Model Performance - **Best Model**: Random Forest - **Recall**: 31.7% (misses 68.3% of actual crises) - **Precision**: 59.1% - **ROC-AUC**: 0.705 - **Accuracy**: 69.4% ### Critical Insight While technically feasible, current models have **limited operational utility** for humanitarian early warning due to low recall rates. The research demonstrates promise but highlights the need for additional data sources and methodological improvements. ## 🗂️ Project Structure