CSHARP Clustering Algorithm is presented for the
purpose of finding clusters of arbitrary shapes and
arbitrary densities in high dimensional feature
spaces. It can find clusters of varying shapes, sizes
and densities; even in the presence of noise and
outliers and in high dimensional spaces as well. It is
based on the idea of letting each data point vote for
its K-nearest neighbors and adopting the points with
the highest votes as clusters' seeds. Two clusters can
be merged if their link strength is sufficient. Any
data point not belonging to a cluster is considered as
noise. The results of our experimental study on
several data sets are encouraging. A wide range of
possible parameters settings yield satisfactory
solutions using the major validation indexes. CSHARP
solutions have been found, in general, superior to
those obtained by DBScan, K-means and Mitosis and
competitive with spectral clustering algorithm.

Download (479 KB) Website Link to this page