Dbscan Clustering Process

Dbscan algorithm

Introduced

Dbscan is a density-based clustering algorithm that includes several important concepts: the core object, the direct density can reach, the density can reach, the density is connected, these several concepts are the layer progressive relations.

Concept

Core object: An object with a neighbor point greater than or equal to MINPT within the ε range, where ε and MINPT are user-defined parameters.

Direct density up to: set P as the core object, |q,p| <= ε, then from P to Q about ε and MINPT direct density can be reached.

Density up to: Set P1,P2,P3....PN as an object sequence, from pi+1 to pi direct density can reach, then pn to P1 density can reach.

Density is connected: from O to P and Q density can be reached, then the density between P and Q is connected.

Clustering process

Iterate through the dataset, starting with a core object p, looking for all the density points of the core object, and adding all the density connections to the cluster where p is located. The core objects of the direct density of p can be found, and the density connection points of these core objects are also added to P, so that recursion is done until it can no longer be expanded.

Dbscan Clustering Process