Public Health Hotspot Allocator

Hotspot Detection (DBSCAN)

Neighborhood radius ε 2.5

Min points to form a core 4

Response Allocation

Mobile response units 5

Response radius r 3.5

Live Metrics

Clusters found (DBSCAN)

–

Mean dist. — Uniform

–

Mean dist. — Hotspot

–

Coverage within r — active mode

–

Incident (hotspot cluster) Incident (noise / sparse) Mobile response unit

How it works

Synthetic overdose/outbreak incidents are scattered across a city grid — some clustered around real hotspots, some background noise. DBSCAN labels each point by counting neighbors within radius ε:

core point:   |N_ε(p)| ≥ minPts
border point: within ε of a core point, but not itself core
noise point:  neither — no dense cluster nearby

N_ε(p) = { q : dist(p, q) ≤ ε }

Clusters grow by connecting core points that are mutual ε-neighbors and absorbing their border points — no need to pre-specify how many clusters exist, and irregularly shaped hotspots are found by density, not distance to a fixed center (unlike k-means).

Each cluster's centroid is the mean position of its members, weighted by cluster size. Two allocation policies for the mobile response units are compared live:

Uniform grid — units spread evenly across the whole area, ignoring where incidents actually happen.
Hotspot-targeted — units go to the largest cluster centroids first; if there are more units than clusters, extras are placed by farthest-point sampling so they still spread out sensibly.

The quality metric is mean response distance — average distance from every incident to its nearest unit — plus the share of incidents within the response radius r. Targeting real hotspots virtually always beats a uniform grid once incidents are non-uniformly distributed, which is the core argument for ML-driven (density-based) resource allocation over fixed coverage patterns in public health logistics.

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.