DBSCAN Clustering Algorithm

DBSCAN Clustering Algorithm

1. Definition

   • Essencial technique
   • Used in domains and industries
   • Acrostic
     • D
       • Density
     • B
       • Based
     • S
       • Spatial
     • C
       • Clustering
     • A
       • Applications
     • N
       • Noise

2. Why we need it?

   • To create arbitrary shapes
     • Varying densities
   • Detects noise

3. Reachability and Connectivity

   • Reachability
     • A data point can be accessed from another data point directly or indirectly
   • Connectivity
     • If two data points belong to the same cluster or not.
   • Two points can be
     • Directly Density-Reachable
     • Density-Reachable
     • Density-Connected

4. Parameter Selection

   • MinPoints
     • Dimensions + 1
   • Epsilon
     • Can be decided from the K-distance graph.

5. Proposed by

   • Martin Ester (1996)

6. Characteristics

   • It's robust to outliers.
   • Parameters
     • Epsilon
       • Radius
     • MinPoints
       • Central point

7. Carlos Daniel Cedeño Ruiz

Images and files in the map