Analytical and predictive modelling project supporting the TOSSIB research programme — a cross-national study of sanitation infrastructure outcomes across Brazilian municipalities.
This project analyses municipal-level data from Brazil to identify the social, political, and infrastructural factors that drive sanitation programme outcomes. The work informed evidence-based planning decisions for the TOSSIB team and contributed to a 14% improvement in targeted programme outcomes.
Core questions:
- Which municipal characteristics (governance, education, investment) most influence sanitation coverage?
- Can we predict which municipalities are at risk of programme underperformance?
- How do outcomes compare across regions and demographic profiles?
| Notebook | Purpose |
|---|---|
TOSSIB.ipynb |
Core EDA — distributions, correlations, feature encoding |
TOSSIB_Analysis.ipynb |
Statistical analysis — chi-square tests, segmentation, regional comparisons |
Tossib new.ipynb |
Predictive modelling — classification models for outcome prediction |
Municipal-level records from the TOSSIB dataset including:
- Sanitation coverage metrics per municipality (
CodMun) - Mayor education level and political affiliation
- Infrastructure investment indicators
- Regional and demographic variables
Exploratory Analysis
- Distribution analysis across sanitation outcome variables
- Chi-square tests for independence between governance features and outcomes
- Correlation analysis across investment and coverage metrics
Feature Engineering
- Ordinal encoding of education levels (Elementary → Higher education)
- Categorical encoding of political and regional variables
- Derived composite indicators
Predictive Modelling
- Binary classification to identify at-risk municipalities
- Models: Logistic Regression, Random Forest, Gradient Boosting
- Evaluation: accuracy, precision, recall, F1
Python · pandas · NumPy · scikit-learn · matplotlib · seaborn · scipy
Findings adopted directly into TOSSIB project team planning decisions. The composite analytical framework contributed to a 14% improvement in targeting efficiency for sanitation interventions across the study municipalities.