Predictive Modelling of Malaria Risk Using the Nigerian Demographic and Health Survey Data

Existing research has not extensively explored the utilisation of machine-learning techniques to predict malaria risk. This study developed a machine-learning model to predict malaria risk based on demographic, environmental and GPS data from the Nigerian Demographic and Health Survey Program (DHS) 2000−2020. The dataset was split into a train (406 covariates) and a test (102 covariates) set. Machine learning algorithms, including Random Forest (RFR), Gradient Boosting (GB), and Logistic Regression (LR), were deployed to accurately predict malaria risk from the dataset. The results indicate that RFR has the lowest MSE (0.0003) and the highest R² (0.9816), making it the model with the best predictive accuracy and optimal for malaria prediction (MalPred) based on the DHS datasets. The regression equation is MalPred = 0.26 − 0.00NC + 0.0053PD − 0.0033TT + 0.00ITN − 0.0070RF + 0.0062MT + 0.10MI − 0.0269TJ − 0.04PET − 0.0115DLS (NC—nightlights composite; PD—population density; TT—travel time; ITNs—insecticide-treated nets; RF—rainfall; MT—minimum temperature; MI—malaria incidence; TJ—temperature in January; PET—potential evapotranspiration; DLS—dry lowland soil). The model showed that MI, MT and PD contributed the most, while TJ and PET contributed the least to malaria risk prediction in Nigeria. This study could be applied to enhancing early predictions of malaria risk using machine learning while also facilitating targeted prevention and allocation of resources in high-risk areas.