mlamborj/comparing-poverty-models

This project contains code for comparing Earth Observation-based machine learning models for estimating poverty patterns with a focus on sub-Saharan African countries. The code was developed as part of my PhD on Spatiotemporal poverty mapping using Earth Observation data and Deep Learning in Africa. # Examining the consistency of Earth Observation-based Machine Learning models for predicting poverty in sub-Saharan Africa This repository contains the code required to reproduce the figures, tables, and supplementary visualizations associated with the manuscript submitted by Mlambo et al. (2026). ## Raw data All the data used in these analyses are available from online repositories and can be downloaded for individual countries. - Demographic and Health Surveys (DHS) can be obtained upon registration. - Relative wealth index (RWI) data by Chi et al.(2022). - International wealth index (IWI) poverty maps by Lee and Braithwaite (2022). - Harmonized wealth index by Yeh et al.’s(2020). ## Instructions Install the required packages from `requirements.txt` into a virtual environment. Download the source data from the links provided and extract into ./data/raw. Run the scripts in ./src/ in order from 1 - 12 to generate the data for figures. Run ./notebooks/Figures.ipynb to reproduce the figures in the article. ## Project Organization ``` ├── data │ ├── external <- Data from third party sources. │ ├── interim <- Intermediate data. │ ├── processed <- The final datasets for reporting and generating figures. │ └── raw <- Folders DHS, Chi, Lee and Yeh containing the source data. │ ├── notebooks <- Jupyter notebooks for generating figures │ ├── reports <- Generated analysis │ └── figures <- Generated figures to be used in the manuscript. │ ├── requirements.txt <- The requirements file for reproducing the analysis environment. │ └── src <- Source code for this project. │ ├── config.py <- Store variables and configuration ``` --------