liveBook · Manning

Border point

As explained in Data Without Labels

A border point is a non-core observation in a DBSCAN cluster that lies within epsilon distance of a core point.

A border point is a non-core observation assigned to a DBSCAN cluster because it lies within ε distance of a core point. section 2.5.2

Use these rules to interpret border points:

  • A border point does not satisfy the minPts density threshold required to be a core point. section 2.5.2
  • A core point can directly reach a border point, but a border point cannot reach another point. section 2.5.2
  • A border point can connect to a cluster through a chain of core points, but it cannot extend that chain. section 2.5.2
  • Points that are neither core nor reachable border points remain outliers and receive no cluster assignment. section 2.5.2

Core-to-border reachability

If X reaches core point Y, and Y lies within ε distance of border point Z, Z belongs to the cluster through Y; travel cannot proceed from Z back to X. section 2.5.2

Figure 2.29 Core points are shown as squares; border points are shown as filled circles, while noise is unfilled circles. Together, these three are the building blocks for DBSCAN clustering.
figure

Compare the visual symbols for core points, border points, and unassigned noise. section 2.5.2

How Data Without Labels covers it

How deep it goes
Defines density status and directional reachability, then stops short of implementation-specific border-point handling.
Examples
Uses Figures 2.29 and 2.30 to show point categories and travel from core points X and Y to border point Z.
What you can do after
Evaluate DBSCAN results by separating core points, reachable border points, and unassigned outliers.
How it connects
Connects border points to DBSCAN, core points, ε-neighborhoods, minPts, density reachability, and outliers.

This explainer was generated by Manning's knowledge graph from the text of Data Without Labels. The section links point to the parts of the book it draws on.