EpiRisk-SSA
A probabilistic epidemic-forecasting research project exploring how epidemiological and environmental data can support earlier awareness of malaria and cholera risk in Sub-Saharan Africa.
Placeholder - Actual dashboard screenshot to be added
Overview
EpiRisk-SSA explores time-series forecasting, risk quantiles, prediction intervals, and outbreak-surge detection. The saved dashboard covers Benin, Ghana, and Nigeria and includes country risk monitoring, model-performance comparisons, calibration views, and outbreak-detection information.
The Problem
Public health systems in Sub-Saharan Africa need earlier awareness of epidemic risk to enable proactive resource allocation and intervention.
Our Approach
Apply time-series forecasting methods to epidemiological and environmental data, producing probabilistic risk forecasts with quantile predictions and outbreak detection.
Methods & Technology
- Time-series forecasting with Temporal Fusion Transformer (TFT)
- Quantile forecasts and pinball loss evaluation
- Split conformal calibration
- Outbreak/surge detection algorithms
- Evaluation against nominal coverage and detection metrics
Evidence & Results
Research dashboard reports 92.2% empirical malaria coverage against a 95% target and 12 of 13 surges detected in the evaluation period. Cholera coverage at 68.6% indicates calibration challenges requiring further work.
Limitations & Considerations
This is a research prototype, not a validated operational public-health service. Coverage metrics are from a defined research evaluation period. Cholera forecasting calibration needs improvement. Clinical deployment would require extensive validation, regulatory review, and integration with public health systems.
Next Steps
Improve cholera forecast calibration. Expand evaluation datasets. Engage public health stakeholders for validation requirements.
Discuss a Similar Project
Working on something similar? Let's talk about what you're building and see if we can help.