Conference Abstract | Volume 8, Abstract NACNDC/19JASH020 (Oral) | Published: 20 Nov 2025
Nebyeye Gift1,&, John Mbazira Ssenkusu1, Simon Kigozi1, Susan Kansiime1, Iving Mumbere1, Iren Wanyana1, Serubungo James1, Batte Charles2, Patrick Katete2, Rawlance Ndejjo1
1Makerere University School of Public Health, Kampala, Uganda, 2Makerere University School of Public Health, Department of Epidemiology and Biostatistics, Kampala, Uganda
&Corresponding author: Nebyeye Gift, Makerere University, School of Public Health. Email: gnebye@gmail.com ORCID: https://orcid.org/0009-0004-2587-6890
Received: 19 Sept 2025, Accepted: 20 Oct 2025, Published: 20 Nov 2025
Domain: Health Informatics
Keywords: Severe Pneumonia; Predictive Model; Machine Learning; XGBoost; Children Under Five; Uganda
©Nebyeye Gift et al. Journal of Interventional Epidemiology and Public Health (ISSN: 2664-2824). This is an Open Access article distributed under the terms of the Creative Commons Attribution International 4.0 License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Cite this article: Nebyeye Gift et al., Development and deployment of a web-based predictive model for weekly severe pneumonia cases among children under five in Kampala using machine learning and ARIMA models. Journal of Interventional Epidemiology and Public Health. 2025;8(ConfProc6):00020. https://doi.org/10.37432/JIEPH-CONFPRO6-00020
Severe pneumonia remains a leading cause of mortality among children under five in Uganda. Proactive public health interventions are hindered by the lack of reliable forecasting tools. This study aimed to develop and deploy a predictive model for weekly severe pneumonia cases in Kampala by integrating meteorological data and influenza-like illness (ILI) indicators.
A time-series study was conducted using 315 weeks of data (2019-2024) from Uganda’s District Health Information Software 2 (DHIS2) and the Uganda National Meteorological Authority. We developed and compared three models: Auto-Regressive Integrated Moving Average with Exogenous variables (ARIMAX), Random Forest, and Extreme Gradient Boosting (XGBoost). Models were trained and validated using time-series cross-validation. The best-performing model was deployed as a web-based application using the Flask framework. Model performance was evaluated using Root Mean Squared Error (RMSE).
The XGBoost model demonstrated the highest predictive accuracy with an RMSE of 85.2, followed by Random Forest (RMSE=92.7) and ARIMAX (RMSE=105.4). An ensemble of XGBoost and ARIMAX further improved performance (RMSE=80.3). Temperature was a statistically significant predictor in the ARIMAX model (p=0.010). The best-performing model was successfully deployed as an accessible web application that allows healthcare workers to input variables like temperature, rainfall, and ILI cases to receive real-time weekly forecasts.
Machine learning models, particularly XGBoost, outperformed traditional statistical models in forecasting pediatric severe pneumonia cases in a low-resource setting. The deployment of a user-friendly, web-based tool enables the practical application of these forecasts for early warning, improving resource allocation, and strengthening public health preparedness in Kampala.
Menu