DengueRadar SG - DAISI ROUND 1 Concept

Inspiration

Dengue cases dropped in 2025, but serotype shifts and incomplete Wolbachia coverage mean outbreaks can return suddenly. We wanted to help NEA move from reactive spraying after clusters form to proactive prevention.

What it does

It forecasts dengue cases nationally 2–4 weeks ahead. It scores neighbourhoods by risk of new clusters. It shows a dashboard with a risk map, top 10 areas to inspect, and explanations for each score.

How we built it

Use open government datasets (cases, clusters, weather, housing, Wolbachia coverage). Build a pipeline on Databricks to ingest, clean, and join data. Train forecasting and classification models. Serve results in a dashboard with natural language Q&A.

Challenges we ran into

Cluster data is only published as live snapshots, so you’ll need to archive it. Thin subzone history may limit granularity. Weather data gaps and alignment issues. Designing a clear end‑to‑end pipeline. Identifying datasets and modelling approach. Framing a solution that directly supports NEA’s decision‑making.

Accomplishments that we're proud of

Designing a clear end‑to‑end pipeline. Identifying datasets and modelling approach. Framing a solution that directly supports NEA’s decision‑making.

What we learned

Outbreak forecasting requires combining epidemiology, environment, and housing data. Usability is as important as accuracy. Open government data can be powerful for public health innovation.