CLIQUE: The CLIQUE Algorithm for Subspace Clustering
Description
The CLIQUE Algorithm finds clusters by first dividing each
dimension into xi equal-width intervals and saving those intervals where the
density is greater than tau as clusters. Then each set of two dimensions is
examined: If there are two intersecting intervals in these two dimensions and
the density in the intersection of these intervals is greater than tau, the
intersection is again saved as a cluster. This is repeated for all sets of
three, four, five,... dimensions. After every step adjacent clusters are
replaced by a joint cluster and in the end all of the clusters are output.
Usage
CLIQUE(data, xi = 10, tau = 0.2)
Arguments
data
A Matrix of input data.
xi
Number of Intervals.
tau
Density Threshold.
References
Rakesh Agrawal, Johannes Gehrke, Dimitrios Gunopulos, and
Prabhakar Raghavan. Automatic Subspace Clustering of High Dimensional
Data for Data Mining Applications. In Proc. ACM SIGMOD, 1999.