Tactical vs. strategic: An adaptable framework for horizon-dependent dengue forecasting using data-driven approaches with serotype and climate covariates in Bangladesh
Aug 2026· PLoS ONE· Vol 21, pp. e0353069 - e0353069· 0 citations· 67 references
Medicine
TL;DR
Results show that dengue forecasting algorithms should be selected according to the intended decision objective and evaluated using task-relevant protocols, and show that MLR showed competitive performance in regime-wise evaluation.
Abstract
Dengue remains a major public health problem in endemic regions, including Bangladesh. Forecasting algorithms relying on climatic variables may not capture epidemiological information such as dengue serotype patterns. This study proposes a horizon-dependent dengue forecasting framework and applies it to Bangladesh. This pipeline included temporal, meteorological, demographic, and epidemiological covariates in a district-panel pipeline and compares classical, machine-learning, and deep-learning algorithms using aggregate, horizon-wise, regime-wise, outbreak-detection, and uncertainty metrics. Coefficient-based interpretation, SHAP, permutation importance, and nonlinear causal-dependence analysis were used to examine predictor impacts across horizons. SARIMAX was used as the baseline model. TFT produced quantile forecasts but showed poor interval calibration during outbreak periods. Results show that no single model performed best across all forecasting horizons: SARIMAX ranked highest for one-step outbreak alerting (precision = 0.886, recall = 0.824, F1 score = 0.854, and ROC-AUC = 0.950), whereas Prophet performed best for pooled magnitude forecasting at horizons 2–6. MLR showed competitive performance in regime-wise evaluation. These findings show that dengue forecasting algorithms should be selected according to the intended decision objective and evaluated using task-relevant protocols.
Background The original district-level dengue forecasting ensemble for Vietnam’s Mekong Delta Region (MDR) showed strong overall accuracy but had two key limitations: forecasted outbreak peaks lagged behind observed timing, and outbreaks were under-detected during periods of rapidly increasing incidence. Methods We evaluated 57 candidate models varying in spatial structure, seasonal representation, and covariates, including flexible harmonic seasonality and climate-driven covariates (mosquito thermal suitability, cumulative precipitation), assessed using the Continuous Ranked Probability Score and Brier score. The six best-performing models were combined into a revised ensemble using district- and horizon-specific Brier score weighting, trained on 2010–2015 data, cross-validated over 2016–2020, and evaluated over 2021–2025 for the on-season period (April–November), using the 95th percentile of historical district-month cases as the epidemic threshold. Results At the 3-month forecast horizon, the revised ensemble achieved 66.3% sensitivity and 89.1% specificity, compared with 43.4% sensitivity and 91.7% specificity for the original model, and overall accuracy improved from 84% to 87% (see Extended Data). The revised model also reduced the temporal lag in outbreak predictions observed previously. Conclusions The revised ensemble improves outbreak detection sensitivity while maintaining high specificity and reduces forecast lag, supporting more timely and targeted dengue prevention and control interventions in the MDR.
Wala Draidi Areed, Vinh Bui, D. M. Weinberger et al.· Wellcome Open Research· 0 citations
Dengue fever poses a major global health threat, with Brazil experiencing severe recurrent outbreaks driven by climatic, socio-economic and mobility factors. Accurate prediction remains challenging owing to dynamically shifting transmission patterns under intervention. This study employs a physics-informed neural network framework to integrate surveillance data with mechanistic models, inferring time-varying intervention parameters. The model demonstrates a strong capacity to fit state-level data from Brazil. Furthermore, it is used to generate 12-week forecasts for future dengue cases. Given the critical influence of climate on transmission dynamics, we also constructed separate eXtreme Gradient Boosting (XGBoost) models for individual states using epidemiological and meteorological data. SHapley Additive exPlanation (SHAP) analysis was applied to identify key climatic drivers and reveal significant regional heterogeneity. Our findings reveal that the drivers of the outbreak exhibit regional heterogeneity, and this geographical variation shows a strong correlation with the intensity of human activities and regional economic development levels. These insights into spatial heterogeneity and predictive modelling can inform the design of more effective, regionally tailored public health strategies for dengue control in Brazil.
Dan-Yang Li, Wei-De Li, Hao-Tian Zhang et al.· Journal of the Royal Society...· 0 citations
TREA-Net is proposed, a Transferable Residual Epidemiological Adaptation Network for dengue forecasting under limited data that improves the corresponding backbone in 9 out of 10 settings, with statistically significant gains.
Dengue is an arboviral disease of high public health relevance, characterized by pronounced temporal variability, nonlinearity, and recurrent outbreaks, which pose challenges to epidemiological surveillance and decision-making. This study evaluated the performance of machine learning methods for short-term forecasting of the weekly dengue morbidity rate in the 27 Brazilian capital cities, comprising the 26 state capitals and Brasília, Federal District, with horizons up to 4 weeks. Epidemiological, climatic, and socioeconomic data were compiled for these capital cities and used to compare a Gated Recurrent Unit (GRU) neural network, formulated as a Multi-Input Multi-Output (MIMO) model, and a Gradient Boosting model (CatBoost), implemented using a Direct forecasting strategy with horizon-specific models. Validation was conducted using a walk-forward approach, with evaluation based on absolute error metrics and the coefficient of determination. The results indicated that the GRU architecture presented recurring limitations, including underfitting, temporal lag, and low capacity to anticipate epidemic peaks. In contrast, the CatBoost model demonstrated greater robustness and better adaptation to the variability of epidemiological time series, showing superior performance in most of the analyzed capitals. The findings reinforce that greater architectural complexity does not necessarily imply better operational performance and highlight the potential of ensemble-based methods for short-term epidemiological surveillance applications. These findings contribute to dengue forecasting by showing that, under a common validation framework, ensemble-based strategies may provide greater operational robustness than recurrent MIMO architectures for short-term prediction in heterogeneous epidemiological settings.
D. C. da Cunha e Silva, L. M. Nery, Nícholas de Paula Nicomedes et al.· International journal of bio...· 0 citations
This study provides the first structured integration of disaster severity covariates in a dengue forecasting framework for Mexico, and results collectively support a differentiated public health response that targets both endemic coastal states and newly affected inland regions.
Hui-Xuan Li, Christopher Lee, Sean Sweeney et al.· Frontiers in Public Health· 0 citations
Accurately forecasting the transmission of infectious diseases, such as Dengue Haemorrhagic Fever (DHF), is essential for robust public health planning. This study assesses the performance of three time-series forecasting models—Exponential Smoothing, Autoregressive Integrated Moving Average (ARIMA), and Seasonal ARIMA (SARIMA)—in predicting annual DHF cases in Malaysia from 2011 to 2021. Model accuracy was evaluated using the Mean Absolute Percentage Error (MAPE). Results indicate that ARIMA and SARIMA yielded the lowest error rates (73.34%), whereas Exponential Smoothing performed less effectively (76.79%). ARIMA was preferred due to its simplicity and comparable predictive power. Although these error rates exceed the standard 10–20% threshold for reliable forecasting, the study demonstrates that these models have significant potential when integrated with more granular data and external covariates. Furthermore, a three-segment piecewise nonlinear regression was employed to identify structural changes in DHF trends. This combined analytical approach provides deeper insights into disease progression and strengthens the evidence base for effective public health interventions.
Farhan Haziq Baderul Hisam, Muhammad Harith Syafiq Zainudin, M. Ismail et al.· International journal of res...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
