christaingabire/Social-Protection-Data-System-Simulation-Rwanda

Synthetic simulation of Rwanda’s social protection data: ETL, analytics, and ML demo for governance # Social Protection Data System Simulation in Rwanda This open-source simulation is inspired by my work at the Rwanda Social Security Board (RSSB). It demonstrates how synthetic data, analytics, and machine learning can be used to explore social protection coverage and inclusion in a transparent, reproducible way. In this project, I'm entirely using without using any confidential data! --- ## Overview - Generates 100k+ rows of fully synthetic citizen-level data (age, gender, income, district, scheme type, etc.) - Cleans and aggregates the data via a lightweight Python ETL pipeline - Produces key indicators: - Enrollment rate by district - Gender gap in coverage - Income-level equity in enrollment - Includes a predictive ML model estimating who’s least likely to be covered - Outputs dynamic charts and metrics for visualization --- ## Interactive Dashboard The data pipeline connects to a lightweight Streamlit dashboard, which allows users to explore coverage and benefit metrics interactively **Features** - Filter by district and scheme type - View key KPIs: enrollment rate, gender gap, benefit distribution - Explore analytics: - Enrollment rate by district - Predicted under-coverage risk (AI model) - Gender composition among enrolled (pie chart) - Enrollment by income level - Export cleaned data as CSV To run locally: ```bash streamlit run src/dashboard_app.py ``` --- ## Folder Structure ``` data/ synthetic_social_protection.csv coverage_metrics.csv src/ generate_data.py cleaning_pipeline.py dashboard_app.py notebooks/ coverage_analysis.ipynb figures/ *.png ``` --- ## Tech Stack `Python` · `Pandas` · `Matplotlib` · `Altair` · `scikit-learn` · `Streamlit` · `SQLite` --- ## Findings & Discussion This simulation reproduces familiar real-world coverage patterns in social protection systems: - Coverage disparities: Some districts show higher simulated enrollment rates, reflecting how geography and infrastructure can influence access - Income-level inequ …