What the study found
The study found that district-level dengue predictions in Bangladesh were strongest when climate, socio-demographic, economic, healthcare, and environmental factors were combined with AI and Bayesian modeling. It also found that climate was the strongest predictor, while poverty and healthcare capacity contributed to prediction.
Why the authors say this matters
The authors conclude that their integrated framework delivers transparent, interpretable predictions and district-level early warnings. They say this supports adaptive dengue outbreak preparedness and resource allocation in Bangladesh.
What the researchers tested
The researchers examined dengue cases across all 64 districts in Bangladesh from 2017 to 2024. They combined Directorate General of Health Services case records with climate, socio-demographic, economic, healthcare, and environmental indicators, and tested Multi-Layer Perceptron, ConvLSTM (Convolutional Long Short-Term Memory), SHAP (Shapley Additive Explanations), and Bayesian spatio-temporal models.
What worked and what didn't
The MLP model achieved the best yearly performance, with accuracy of 0.93 and ROC-AUC of 0.99. ConvLSTM performed best for monthly prediction, with recall of 0.88 and ROC-AUC of 0.81, and Bayesian BYM2_RW2 with lagged effects improved predictive fit with DIC of 3671.055. The abstract does not report a direct comparison showing which approaches did not work well beyond these performance differences.
What to keep in mind
The abstract does not describe detailed limitations of the study. It also reports that the findings come from Bangladesh districts and the 2017 to 2024 period, so the stated scope is specific to that setting and dataset.
Key points
- The study analyzed dengue cases across all 64 districts in Bangladesh from 2017 to 2024.
- Climate was the strongest predictor of dengue transmission in the models.
- Poverty and healthcare capacity also contributed to dengue prediction.
- The MLP model had the best yearly performance, with accuracy of 0.93 and ROC-AUC of 0.99.
- ConvLSTM was best for monthly prediction, with recall of 0.88 and ROC-AUC of 0.81.
Disclosure
- Research title:
- District-level dengue prediction improved with hybrid AI and Bayesian models
- Authors:
- Md. Abu Bokkor Shiddik, Farzana Zannat Toshi, Sadia Yesmin, S. M. Mahfujar Rahman
- Institutions:
- Begum Rokeya University, Begum Rokeya University, Begum Rokeya University, Begum Rokeya University
- Publication date:
- 2026-03-05
- OpenAlex record:
- View
- Image credit:
- Muhammad Mahdi Karim, Wikimedia Commons, GFDL 1.2
Get the weekly research newsletter
Stay current with scholarly research without reading academic papers — one filtered digest, every Friday.