jerzybala/rural-mhq-xgboost-tanzania

This experiment is part of a study to understand lifestyle drivers of MHQ (Mental Health Quotient) in Tanzania. The analysis explores XGBoost+SHAP methods. # XGBoost Analysis of Tanzanian MHQ Data ## Project Background This project investigates the lifestyle drivers of Mental Health Quotient (MHQ) in rural Tanzania using machine learning. XGBoost and SHAP analysis is used to identify and interpret key predictors of mental health outcomes,supporting group and variant comparisons (Rural.Hadza versus GM in 24 in age segments with/without use of Smart Phones). ## Main Script - **xgb_performance_table_generator.py** - Loops through all group/variant combinations - Applies one-hot encoding and preprocessing - Runs 5-fold cross-validation for each combination - Computes and exports: - Summary metrics (mean/std for AUC, accuracy, F1, precision, recall, specificity, balanced accuracy, MCC, Cohen’s kappa) to `xgb_performance_table.xlsx` - Per-fold metrics to `xgb_performance_table_per_fold.xlsx` - Confusion matrices for each fold to `xgb_confusion_matrices.xlsx` - Feature importances (mean/std across folds) to `xgb_feature_importances.xlsx` ## Data - Input: `rural_gmdata_forML.csv` - Target: Binary classification (Succeeding vs. Struggling, based on Overall.MHQ) ## Usage 1. Activate your Python environment: ```bash source /path/to/venv/bin/activate ``` 2. Run the main script: ```bash python xgb_performance_table_generator.py ``` 3. Results will be saved as Excel files in the same directory. ## Requirements - Python 3.8+ - pandas, numpy, scikit-learn, xgboost, openpyxl Install requirements (if needed): ```bash pip install pandas numpy scikit-learn xgboost openpyxl ``` ## Output Files - `xgb_performance_table.xlsx`: Summary metrics (mean/std) for each group/variant - `xgb_performance_table_per_fold.xlsx`: All per-fold metrics - `xgb_confusion_matrices.xlsx`: Confusion matrices for each fold - `xgb_feature_importances.xlsx`: Feature importances (mean/std across folds) ## Customization - Edit `VARIANTS` and `GROUPS` in the script to change analysis scope. - Adjust model parameters or metrics as needed. ## Contributors - …