While LSTM networks demonstrate promising performance in modeling nonlinear temporal dynamics, the ARIMA model provided the most accurate predictions for the studied dengue dataset, suggesting that while deep learning models are capable of capturing complex temporal patterns, their advantage over well-tuned statistical models is not guaranteed for this dataset.
Abstract
Accurate forecasting of infectious disease dynamics is essential for effective public health planning and early outbreak response. Traditional statistical models and modern deep learning approaches have both been widely used for epidemiological time–series forecasting. Dengue fever remains a major public health concern in Bangladesh, where seasonal outbreaks often produce sharp and unpredictable infection spikes. Reliable forecasting models are therefore important for anticipating infection trends and supporting early intervention strategies. This study investigates the performance of Long Short-Term Memory (LSTM) networks for predicting daily dengue infection counts in Bangladesh using surveillance data from 2010 to 2022. Multiple LSTM architectures were developed by varying the number of LSTM units, lookback window size, and data preprocessing strategies. A fixed time-based data split was used, with training data up to 30 September 2021 and testing on subsequent observations to avoid information leakage. The proposed LSTM models were compared with several baseline forecasting approaches, including Naive, Seasonal-Naive, and ARIMA models. Model performance was evaluated using standard forecasting metrics including Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Symmetric Mean Absolute Percentage Error (SMAPE). Experimental results indicate that classical statistical models remain strong benchmarks for dengue forecasting. Among all evaluated models, the ARIMA(2,1,2) model achieved the best predictive performance on the test dataset with MAE of 8.40 and RMSE of 14.72. The best-performing LSTM configuration produced competitive results with MAE of 8.92 and RMSE of 15.87, outperforming simple baseline approaches such as the Naive and Seasonal-Naive models. These findings suggest that while deep learning models are capable of capturing complex temporal patterns, their advantage over well-tuned statistical models is not guaranteed for this dataset. The results highlight the importance of benchmarking deep learning models against classical time-series approaches when forecasting epidemiological data. While LSTM networks demonstrate promising performance in modeling nonlinear temporal dynamics, the ARIMA model provided the most accurate predictions for the studied dengue dataset.
Accurate multi-week dengue forecasting supports timely vector-control interventions, outbreak preparedness, and healthcare resource allocation. However, newly established surveillance systems often lack the historical data needed to train reliable neural forecasting models. Although pretrained time-series models offer promising zero-shot forecasts, their cross-domain training may not capture local epidemiological dynamics. We propose TREA-Net, a Transferable Residual Epidemiological Adaptation Network for dengue forecasting under limited data. TREA-Net augments neural forecasting backbones with projections from an Environmental Time-Series Susceptible-Infected-Recovered model and learns a lightweight gated residual correction transferable from data-rich to data-scarce regions. Its node-invariant design accommodates surveillance systems with different numbers of locations, while target adaptation requires learning only two global parameters. We transfer knowledge from long-running dengue surveillance in Colombia and Nicaragua to 8-week-ahead forecasting in Mexico and Malaysia using only 78 or 104 weeks of target data. Across five neural backbones and ten transfer settings, TREA-Net improves the corresponding backbone in 9 out of 10 settings, with statistically significant gains. When integrated with TiRex, a foundation model for forecasting, it achieves the lowest mean absolute error across all target datasets. Conformal prediction further maintains empirical coverage while reducing 8-week prediction-interval width by 29.6% in Mexico. These results demonstrate TREA-Net's potential as a lightweight and portable early-warning framework for health agencies with limited surveillance data.
Inesh Shukla, Madhurima Panja, Tanujit Chakraborty et al.· 0 citations
A comprehensive review of CI models for outbreak prediction, comparing supervised and unsupervised methods such as Support Vector Machines (SVM), Random Forest (RF), Artificial Neural Networks (ANN), Long Short-Term Memory (LSTM), and hybrid models.
Z. Abdullahi· International Journal of App...· 0 citations
Findings indicate that SARIMAX is more suitable for forecasting dengue incidence characterized by strong seasonal patterns and relatively limited observations, than LSTM.
George Elmar, Asriyanik, Winda Apriandari· Kontribusia (Research Dissem...· 0 citations
Accurate forecasting of influenza-like illness (ILI) incidence is essential for public health surveillance and early epidemic intervention. Traditional statistical models effectively capture linear temporal trends but often fail to model the complex nonlinear dynamics of disease transmission. To address this limitation, this study proposes a hybrid forecasting framework that integrates Autoregressive Integrated Moving Average (ARIMA), Long Short-Term Memory (LSTM), temporal attention mechanisms, and Transformer architecture. Weekly ILI surveillance data collected from Xinjiang and Shandong provinces, China, between 2023 and 2025 were used for model development and validation. Meteorological variables, including weekly average temperature and relative humidity, were incorporated as exogenous features. The proposed framework decomposes the forecasting task into linear and nonlinear components, where ARIMA captures long-term linear trends and an attention-enhanced LSTM–Transformer branch models nonlinear temporal dependencies. Experimental results demonstrated that the proposed model achieved superior predictive performance on the Xinjiang Production and Construction Corps dataset, with an MAE of 29.371, RMSE of 41.334, MAPE of 10.744%, and R2 of 0.579. Diebold–Mariano tests confirmed statistically significant improvements over standalone ARIMA, LSTM, and ARIMA– LSTM models (P < 0.05). Furthermore, cross-regional validation on the Yantai High-Tech Zone dataset showed comparable forecasting accuracy, indicating strong generalization capability under different climatic conditions. These findings suggest that the proposed hybrid framework effectively captures both linear and nonlinear transmission patterns of ILI and provides a practical tool for early warning and public health resource allocation in grassroots disease surveillance systems.
Jing-Long Chen, Li-Min Xu, Q. Yuan et al.· 2026 8th International Confe...· 0 citations
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
The development and assessment of a machine learning-driven early warning system for infectious disease prediction using geospatial big data from South-Western Nigeria, at the level of Local Government Area outperform conventional surveillance systems in developing countries.
I. Adewumi, N. Bakare, W. Ajayi et al.· London Journal of Physics· 0 citations