KwabenaAsabere/ghana-malaria-spatial-ml

Spatial machine learning of malaria prevalence in Ghana using DHS data, environmental covariates, Random Forest, and spatial cross-validation. # Spatial Machine Learning of Malaria Prevalence in Ghana ## Project Overview This project applies **spatial machine learning** to model and map malaria prevalence across Ghana, using the **2022 Ghana Demographic and Health Survey (GDHS)** as the primary source of malaria prevalence data. Georeferenced survey clusters from the **2022 GDHS** are integrated with environmental, demographic, land-cover, and accessibility-related geospatial covariates. A **Random Forest regression model** is trained using spatial block cross-validation and subsequently applied to raster predictors to generate continuous estimates of malaria prevalence across Ghana. The project demonstrates how nationally representative household survey data can be combined with **GIS, remote sensing, spatial epidemiology, and machine learning** to support subnational malaria risk assessment and geographic targeting of public health interventions. --- ## Objectives The main objectives are to: - Examine geographic variation in malaria prevalence reported in the **2022 Ghana Demographic and Health Survey**. - Integrate **2022 GDHS malaria prevalence and georeferenced cluster data** with environmental and geographic covariates. - Develop a Random Forest model for predicting malaria prevalence across Ghana. - Use **spatial cross-validation** to evaluate geographic generalizability. - Identify important environmental and geographic predictors of malaria prevalence. - Generate continuous malaria prevalence estimates for locations not directly sampled by the 2022 GDHS. --- ## Data Source ### 2022 Ghana Demographic and Health Survey (GDHS) The primary epidemiological data source for this project is the **2022 Ghana Demographic and Health Survey (GDHS)**. The GDHS provides nationally representative health and demographic information collected from households across Ghana. For this analysis, malaria prevalence estimates are linked to the geographic locations of DHS survey clusters. The primary outcome …