jprmaulion/cholera-gedeo-ethiopia-ml-modeling

Predictive modeling of cholera severity in the Gedeo Zone, Ethiopia, using epidemiological and clinical data. This notebook applies machine learning techniques, including feature engineering, Bayesian hyperparameter optimization, and classification with XGBoost and Random Forest. # Integration of Spatial Patterns and Machine Learning for Cholera Risk Assessment in Southern Ethiopia (PART 2) ## Predictive Modeling of Cholera Severity in Gedeo Zone, Ethiopia Using Machine Learning This repository contains a Jupyter notebook that implements a machine learning pipeline to predict cholera severity based on individual-level epidemiological and clinical data from the Gedeo Zone. The notebook covers feature engineering, data preprocessing, hyperparameter tuning using Bayesian optimization, and model evaluation with XGBoost and Random Forest classifiers. The goal is to develop predictive models that support targeted decision-making and improve surveillance efforts in cholera-affected settings. ## Overview The notebook covers: - Feature engineering from epidemiological and clinical cholera case data, including temporal and environmental variables. - Data preprocessing techniques, such as encoding categorical variables and scaling numerical features. - Handling class imbalance through the calculation of weighting parameters for model training. - Hyperparameter tuning of XGBoost using Bayesian optimization with cross-validation to improve the predictive performance. - Development and evaluation of predictive models, including XGBoost and Random Forest classifiers. - Model performance assessment using metrics such as accuracy, ROC AUC, precision, recall, F1-score, and confusion matrices. - Discussion of the model’s strengths, limitations, and potential applications in clinical decision support and public health surveillance. This notebook enhances the analysis by developing data-driven solutions to identify severe cholera cases, thereby complementing the spatial and descriptive insights obtained from the previous exploratory phase. ## File Structure - `ml_gedeo_cholera_f.ipynb` — Jupyter notebook containing the complete machine learning workflow for predicting cholera severity, including feature engineering, model training, hyperparameter tun …