Prediction of Malaria Incidence using Machine Learning Models in the Presence of Climatic Factors in Kebbi State, Nigeria

Abstract

Malaria remains a critical public health challenge in Nigeria, with Kebbi State experiencing significant transmission burdens influenced by climatic variability. Traditional statistical methods, while useful for identifying associations, often fail to capture the complex, nonlinear, and delayed effects of climatic variables on malaria transmission. This study aimed to assess the impact of key climatic variables on malaria incidence and develop predictive models using machine learning approaches, specifically Random Forest (RF) and Artificial Neural Networks (ANN), in Kebbi State, Nigeria. Monthly malaria incidence data from 2014 to 2022 across all 21 Local Government Areas in Kebbi State were combined with corresponding meteorological records from the Nigerian Meteorological Agency. Preprocessing included normalization, temporal feature engineering (lagged and rolling means), and creation of both regression and binary classification targets. Model performance was evaluated on an independent test set (20% of data) using R², RMSE, MAE for regression, and accuracy, precision, recall, and F1-score for classification. RF regression achieved R² = 0.997, RMSE = 110.93, and MAE = 37.52 on the test set, while ANN achieved R² = 0.991, RMSE = 178.43, and MAE = 84.57. For classification, RF achieved 89.5% test set accuracy with balanced precision and recall, whereas ANN achieved 89.1% test set accuracy. Feature importance analysis confirmed lagged rainfall and humidity as the most influential predictors. Traditional statistical models (MLR) showed limited explanatory power (R² = 0.068-0.097), justifying the machine learning approach. Machine learning models, particularly Random Forest, effectively capture the nonlinear, time-dependent relationships between climate and malaria transmission, significantly outperforming conventional statistical approaches. This study provides a robust predictive framework for malaria surveillance and early warning systems in Kebbi State, highlighting the importance of integrating climatic and temporal predictors for improved disease forecasting.