Socioeconomic Determinants of Suicide Rates - ML Analysis

Machine learning analysis to identify the top 5 socioeconomic determinants of suicide rates using World Bank (WDI) and WHO data. Systematic reduction from 1509 → 5 interpretable variables focused on mental health public policy.

📄 Read the full paper | 📊 View results

🎯 Main Results

Decision Tree (best model): R² = 0.82, MSE = 8.56
Lasso Regression: R² = 0.24, MSE = 36.82

Top 5 Identified Determinants

Variable	Importance	Effect	Interpretation
🚺 Female labor force participation	0.296	⬆️	"Double burden" stress
🏙️ Population density	0.284	⬇️	Access to services/social networks
🏭 Industrial employment	0.169	⬆️	Adverse working conditions
⚡ Access to electricity	0.137	⬇️	Development indicator
🏥 Private health spending	0.115	⬇️	Access to mental health

📊 About the Project

Analysis of 185 countries (2000-2021) integrating World Bank and WHO data to identify socioeconomic factors that most influence suicide rates. Complete pipeline: data collection → variable selection → modeling → interpretation.

Why this matters? Suicide affects 720+ thousand people/year (WHO). Identifying actionable determinants helps develop effective public policies (aligned with UN SDGs).

Methodology Summary

1. Data: 1509 WDI variables + age-standardized suicide rates (WHO)
2. Selection: Correlation filters + Maximum Independent Set → 74 variables
3. Modeling: Decision Trees + Lasso Regression (5-fold CV)
4. Interpretation: Importance analysis + direction of associations

� Quick Start

# Install requirements
pip install -r requirements.txt

# Run best model (Decision Tree with 5 key variables)
python scripts/modeling/generic_regression_crossval.py --model decision_tree --mode interpretable

# Generate comparison plots
python scripts/plot_comparacao_modelos.py

📁 Structure

├── data/
│   ├── raw/          # Original API data
│   ├── interim/      # Intermediate processing  
│   └── processed/    # Final datasets
├── scripts/
│   ├── data_processing/  # Extraction and processing
│   └── modeling/         # ML models
├── results/          # Model outputs
└── src/             # Utilities and feature selection

🔧 Requirements

pip install -r requirements.txt

Main dependencies: Python 3.8+, scikit-learn, pandas, numpy, matplotlib, seaborn, requests

🎯 Contributions

• Systematic reduction 1509→5 interpretable variables
• Directionality analysis (decision trees)
• Reproducible pipeline for epidemiological studies
• Actionable insights for suicide prevention

⚠️ Limitations

• Aggregated data (ecological fallacy)
• Imputation required for all observations
• Directionality analysis could be improved (SHAP/LIME)

👥 Authors

Luan Pereira Pinheiro and Sofia Leopoldo - University of São Paulo

📄 Full Article: PDF | 📁 Data & Code: Available in this repository

---

This research contributes to understanding socioeconomic factors in suicide prevention through interpretable ML, aligned with UN SDGs.