cogaba/Machine-learning-to-guide-HIV-Interventions

R Code for predicting HIV status using random forest and XG Boost on data from four countries in Sub-Saharan Africa # Machine-learning-to-guide-HIV-Interventions R Code for predicting HIV status using random forest and XG Boost on data from four countries in Sub-Saharan Africa # Machine Learning to Guide HIV Interventions This repository contains the analysis code used in the study: **“Leveraging Modern Machine Learning Techniques to Guide Targeted HIV Interventions and Prevention Strategies in Sub-Saharan Africa .”** The scripts implement statistical and machine learning methods to identify predictors of HIV positivity and explore cross-country patterns using population-based HIV survey data. The analysis supports evidence-based targeting of HIV prevention and treatment interventions. --- # Repository Structure The repository includes the following R scripts: ### 1. SMOTE Random Forest Model **File:** smote_random_forest_model.R Purpose: * Applies **SMOTE (Synthetic Minority Oversampling Technique)** to address class imbalance in HIV status. * Trains a **Random Forest model** to predict HIV positivity. * Extracts variable importance to identify key predictors. --- ### 2. XGBoost Model **File:** xgboost_model.R Purpose: * Implements an **Extreme Gradient Boosting (XGBoost)** model. * Provides an alternative machine learning approach for predicting HIV status. * Allows comparison of model performance with Random Forest. --- ### 3. Logistic Regression Analysis **File:** logistic_regression_analysis.R Purpose: * Fits a **logistic regression model** to estimate associations between predictors and HIV status. * Generates plots to aid interpretation and comparison with machine learning models. --- ### 4. Chi-Square Analysis **File:** chi_square_analysis.R Purpose: * Performs **chi-square tests** to examine bivariate associations between categorical predictors and HIV status. --- ### 5. Rank Correlation and Heatmap **File:** kendall_spearman_heatmap.R Purpose: * Calculates **Spearman and Kendall rank correlations**. * Produces **heatmaps** to compare predic …