Multiscale forecasting of dengue alerts from notifications, historical seasonality, and socio-spatial vulnerability

• Silmara Barnabé UFRGS
• Maximus Borges da Rosa UFRGS
• Guilherme Malta UFRGS
• Weverton Luis da Costa Cordeiro UFRGS
• Carla M. D. S. Freitas UFRGS
• Mariana Recamonde-Mendoza UFRGS / HCPA

DOI: https://doi.org/10.5753/eniac.2026.30172

Resumo

Suspected dengue notifications precede laboratory confirmation and are an early proxy for healthcare demand. We frame surveillance as alert prioritization: ranking municipality-weeks by the risk of high notification burden over the next four weeks, with the alert set per municipality at the 80th percentile of training burden. Following CRISP-DM, we build a multiscale weekly dataset of notification history, acceleration, seasonal anomaly, climate, and socio-spatial vulnerability using training data from 2010 to 2024. On the 2025 out-of-time test, the best classifier reached an Average Precision (AP) of 0.912 and precision@top20% of 0.941, against 0.837 for a strong seasonal baseline; the gain held in 2024–2025 (AP 0.938 vs. 0.841), while finer UDH/locality granularity did not improve on the municipal model.

Referências

Aleixo, R., Kon, F., Rocha, R., et al. (2022). Predicting dengue outbreaks with explainable machine learning. In Workshop on Artificial Intelligence for Health (AI4Health).

Barcellos, C. and Lowe, R. (2014). Expansion of the dengue transmission area in Brazil: The role of climate and cities. Trop. Med. Int. Health, 19(2):159–168.

Barnabé, S., Rosa, M. B. d., Freitas, C. M. D. S., et al. (2025). Previsão de surtos de dengue em nível de unidade de desenvolvimento humano (UDH) no Brasil: Uma abordagem integrada e explicável. In Anais da I Escola Regional de Aprendizado de Máquina e Inteligência Artificial da Região Sul. 1The curated dataset and the preparation pipeline are available at [link].

Borges, M. E., Codeço, C. T., Machado, D., and Almeida, A. (2026). Epidemiological and digital syndromic surveillance data on dengue, chikungunya, and SARI in Brazil. Scientific Data, 13:68.

Cardim, L. L., Teixeira, M. d. G., Costa, M. d. C. N., Mitka, W. M., Teixeira, C. S. S., Santos, A. C., Santos, G., Souza, A. P. F. d., Smeeth, L., Barreto, M. L., Brickley, E. B., Paixao, E. S., and Pescarini, J. M. (2025). Socioeconomic markers of dengue mortality in the 100 Million Brazilian Cohort (2007–2018): A nationwide registry-based cohort study. PLOS Negl. Trop. Dis., 19(11):1–16.

Chen, X. and Moraga, P. (2025). Forecasting dengue across brazil with lstm neural networks and shap-driven lagged climate and spatial effects. BMC Public Health, 25(1).

Chen, X. and Moraga, P. (2026). Dengue forecasting and outbreak detection in Brazil using LSTM: Integrating human mobility and climate factors. Infect. Dis. Model., 11(1):338–354.

da Silva Neto, S. R., de Oliveira, T. T., Teixeira, I. V., et al. (2022). Arboviral disease record data – dengue and chikungunya, Brazil, 2013–2020. Scientific Data, 9:198.

de Oliveira, F. S., Santos, D. E., Bruhn, F. R. P., et al. (2025). Forecasting dengue outbreaks with machine learning: A case study in Porto Alegre. In Anais da I Escola Regional de Aprendizado de Máquina e Inteligência Artificial da Região Sul. SBC.

Fernandes, M. S., Cordeiro, W., and Recamonde-Mendoza, M. (2021). Detecting aedes aegypti mosquitoes through audio classification with convolutional neural networks. Comput. Biol. Med., 129:104152.

Friedman, J. H. (2001). Greedy function approximation: A gradient boosting machine. The Annals of Statistics, 29(5):1189–1232.

Islam, J., Frentiu, F. D., Devine, G. J., et al. (2025). A state-of-the-science review of long-term predictions of climate change impacts on dengue transmission risk. Environ. Health Perspect., 133(5):056002.

Ke, G., Meng, Q., Finley, T., et al. (2017). LightGBM: A highly efficient gradient boosting decision tree. In NeurIPS, volume 30, pages 3146–3154.

Lee, S. A., Economou, T., Catão, R. d. C., et al. (2021). The impact of climate suitability, urbanisation, and connectivity on the expansion of dengue in 21st century Brazil. PLOS Negl. Trop. Dis., 15(12):e0009773.

Lowe, R., Bailey, T. C., Stephenson, D. B., et al. (2011). Spatio-temporal modelling of climate-sensitive disease risk: Towards an early warning system for dengue in Brazil. Computers & Geosciences, 37(3):371–381.

Lowe, R., Bailey, T. C., Stephenson, D. B., et al. (2013). The development of an early warning system for climate-sensitive disease risk with a focus on dengue epidemics in Southeast Brazil. Statistics in Medicine, 32(5):864–883.

Lowe, R., Coelho, C. A. S., Barcellos, C., et al. (2016). Evaluating probabilistic dengue risk forecasts from a prototype early warning system for Brazil. eLife, 5:e11285.

Lowe, R., Lee, S. A., O’Reilly, K. M., et al. (2021). Combined effects of hydrometeorological hazards and urbanisation on dengue risk in Brazil: A spatiotemporal modelling study. The Lancet Planetary Health, 5(4):e209–e219.

Ong, S. Q., Isawasan, P., Ngesom, A. M. M., et al. (2023). Predicting dengue transmission rates by comparing different machine learning models with vector indices and meteorological data. Scientific Reports, 13(1):19129.

Palmer, D. Q., Motta, M., Moura, E., et al. (2026). Dengue hospitalizations in Brazil: Forecasting with climatic and physicians’ digital search data under real-world reporting delays. PLOS Digital Health, 5(5):e0001206.

Pedregosa, F., Varoquaux, G., Gramfort, A., et al. (2011). Scikit-learn: Machine learning in Python. J. Mach. Learn. Res., 12:2825–2830.

Rosa, M. B., Tavares, G. C., and Recamonde-Mendoza, M. (2026). Low-cost dengue triage: Predicting disease severity using machine learning without laboratory biomarkers. In Anais do XXVI Simpósio Brasileiro de Computação Aplicada à Saúde. SBC.

Salim, M. F., Satoto, T. B. T., Daniel, D., et al. (2024). Digital health interventions in dengue surveillance to detect and predict outbreak: a scoping review. Open Public Health J., 17(1):e18749445283264.

Siraj, A. S., Oidtman, R. J., Huber, J. H., et al. (2017). Temperature modulates dengue virus epidemic growth rates through its effects on reproduction numbers and generation intervals. PLOS Negl. Trop. Dis., 11(7):e0005797.

Wilder-Smith, A., Gubler, D. J., Weaver, S. C., et al. (2017). Epidemic arboviral diseases: priorities for research and public health. Lancet Infect. Dis., 17(3):e101–e106.

Wiratmadja, I. I., Salamah, S. Y., and Govindaraju, R. (2018). Healthcare data mining: Predicting hospital length of stay of dengue patients. J. Eng. Technol. Sci., 50(1):110–126.

Wirth, R. and Hipp, J. (2000). CRISP-DM: Towards a standard process model for data mining. In Proc. 4th Int. Conf. Practical Application of Knowledge Discovery and Data Mining, pages 29–39. Springer.

Zhang, W.-X., Zhao, T.-Y., Wang, C.-C., He, Y., Lu, H.-Z., Zhang, H.-T., Wang, L.-M., Zhang, M., Li, C.-X., and Deng, S.-Q. (2025). Assessing the global dengue burden: Incidence, mortality, and disability trends over three decades. PLOS Negl. Trop. Dis., 19(3):1–18.