Data analysis project on banking access, trends & ML for financial inclusion in South Africa
π Financial Inclusion in South Africa
This project investigates financial inclusion trends in South Africa using open financial datasets. It includes data cleaning, analysis, dashboards, and a machine learning model to predict loan default risk.
π§° Tools & Technologies
Python (Pandas, Seaborn, Scikit-learn)
Power BI (interactive dashboard)
Tableau (storytelling dashboard)
Jupyter Notebook
Open financial datasets (Global Findex, SARB, Stats SA)
π― Project Objectives
Understand disparities in access to banking & credit
Identify socio-economic factors behind exclusion
Predict likelihood of loan default using ML
Present findings through clear visual dashboards
π Project Structure
bash
Copy
Edit
financial-inclusion/
βββ data/
β βββ cleaned_financial_data.csv
βββ notebooks/
β βββ financial_analysis.ipynb
βββ dashboards/
β βββ powerbi_dashboard.pbix
β βββ tableau_dashboard.twbx
βββ model/
βββ loan_default_model.py
π§Ό Data Cleaning
Performed in Python:
Filled missing income/employment values
Removed duplicates & irrelevant columns
Categorical encoding and normalization
π Power BI Dashboard
Key insights:
% unbanked by age & gender
Credit access vs. income level
Map of underserved provinces
π‘ Screenshot goes here (add .png file to dashboards/ folder)
π Tableau Dashboard
A storytelling dashboard highlighting:
How financial access has changed over time
Correlation between education and credit access
π‘ Add your Tableau screenshots or embed GIF previews
π€ ML Model β Loan Default Prediction
Python model using:
Logistic Regression & Decision Trees
Features: income, employment, age, credit history
Output: likelihood of default
bash
Copy
Edit
# Run model
python model/loan_default_model.py
Result: ~82% accuracy on test data
π¬ Key Findings
Youth (18β25) and informal workers are underbanked
Education strongly correlates with financial access
Default risk is higher among low-income, unemployed users
πββοΈ Author
Ivy Sadiki
Senior Test Analyst | Data & A β¦