Neural networks for dengue forecasting: a systematic review

TL;DR

This systematic review assesses how neural networks have been applied to forecasting dengue incidence across diverse settings. It finds that shallow ANNs predominate, with meteorological and epidemiological inputs, and that about half of studies with comparators report NN superiority, though architectures and evaluation practices vary widely. The review identifies opportunities in using deeper networks, richer input features, transfer learning, and improved transparency to enhance forecast utility for public health decision-making. It urges standardized benchmarking and open-data sharing to enable robust cross-context comparisons and policymaker trust in neural-network–based dengue forecasting.

Abstract

Background: Early forecasts of dengue are an important tool for disease mitigation. Neural networks are powerful predictive models that have made contributions to many areas of public health. In this study, we reviewed the application of neural networks in the dengue forecasting literature, with the objective of informing model design for future work. Methods: Following PRISMA guidelines, we conducted a systematic search of studies that use neural networks to forecast dengue in human populations. We summarized the relative performance of neural networks and comparator models, architectures and hyper-parameters, choices of input features, geographic spread, and model transparency. Results: Sixty two papers were included. Most studies implemented shallow feed-forward neural networks, using historical dengue incidence and climate variables. Prediction horizons varied greatly, as did the model selection and evaluation approach. Building on the strengths of neural networks, most studies used granular observations at the city level, or on its subdivisions, while also commonly employing weekly data. Performance of neural networks relative to comparators, such as tree-based supervised models, varied across study contexts, and we found that 63% of all studies do include at least one such model as a baseline, and in those cases about half of the studies report neural networks as the best performing model. Conclusions: The studies suggest that neural networks can provide competitive forecasts for dengue, and can reliably be included in the set of candidate models for future dengue prediction efforts. The use of deep networks is relatively unexplored but offers promising avenues for further research, as does the use of a broader set of input features and prediction in light of structural changes in the data generation mechanism.

Figures (10)

• Figure 1: Auxiliary variables to dengue epidemiological data included in the selected studies, displaying the frequency with which studies included various categories of predictor variables and their combinations. The top bars measure their intersection size in the selected papers, and the central table shows the interplay between multiple variables. 'Meteorologic' data is present in most (69%) studies.
• Figure 2: Geographic and temporal distribution, with green bars highlighting the most frequent category in each case. The left panel shows the frequency with which studies predict dengue cases at the city, district, province, and country level. The right panel shows the frequency with which studies predict dengue cases at daily, weekly, monthly, seasonal, and yearly level. The "entire period" of dengue data accounts for total number of cases in a certain period, and was used by studies that predicted dengue fever risk in specific regions tab11_Anderson2018tab43_Andersson2019.
• Figure 3: Geographic coverage. a) Number of studies predicting dengue in each country from 2008 to 2025 and b) number of studies predicting dengue in different geographic regions, by year of publication.
• Figure 4: Accessibility of studies. The vertical bars show the frequency of open and restricted access, publication venue, data availability, and source code availability of the selected studies. The gray flows indicate how frequently combinations of characteristics appear together.
• Figure 5: Types of neural network architectures employed by the selected studies and their respective publication year, grouped by structural similarity. The green bars indicate the most commonly used architecture, which was the artificial neural network (ANN).