Home

Welcome to the improved-dbscan wiki! density-based clustering algorithm in Python

Works with 2D points. Usage

Run with parameters (optional):

-f filename.csv - path to dataset file with 2D points (default "datasets/annulus.csv", semicolon separated values)
-eps 5 - epsilon parameter (see more (wiki)[https://en.wikipedia.org/wiki/DBSCAN#Original_Query-based_Algorithm]) (default 5)
-min 10 - minimum number of points required to form a dense region parameter (default 10)

Example:

python dblp.py -f datasets/annulus.csv -eps 5 -min 10

Output

Silhouette Coefficients
SSE
Dunn index
Size of clusters
Python plot with points, colored by clusters

Improvement

Better CVI Value than Original DBScan
Introduced Eps' to reduce the area to pick next point
Core Point searches for next point to pick in region between Eps and Eps'