This project is part of Kaggle's Data Science for Good initiative, where the goal is to support Kiva.org, a nonprofit microlending platform, by estimating the poverty levels of loan recipients around the world. By combining Kiva’s internal data with external datasets, I aim to uncover regional insights that help guide better lending strategies and deepen Kiva’s social impact.
Kiva wants to better understand the welfare levels of borrowers in the regions it serves. The objective is to pair Kiva's loan data with external sources to estimate poverty at a granular geographic level.
Strong solutions will:
- Map loan and borrower features to regional poverty indicators
- Accurately geocode borrower locations (even when vaguely described)
- Disaggregate poverty data by gender, loan sector, or borrowing behavior
- Provide insights into how Kiva’s lending decisions relate to local economic conditions
Kiva provides a rich dataset covering:
- Loan details (amount, use case, sector, terms)
- Borrower demographics (gender, group size)
- Location info (village, country, etc.)
- Field partner metadata
⚠️ Note: Due to GitHub’s file size limits, the full dataset is not included in this repository.
You must download it directly from Kaggle’s challenge page.
To download via CLI:
kaggle competitions download -c data-science-for-good-kiva-crowdfunding