A Novel Hierarchical Clustering Algorithm Based on Density Peaks for Complex Datasets

Algorithm 1

HCDP main steps

Require: dataset , neighborhood , child threshold .

1: Compute -nearest density and distance for each point

2: Generate a tree by connecting from one point to its nearest point with higher density, and assign the whole tree as a single cluster.

3: Sort all the edges of the tree with respect to the weights in descending order.

4: repeat

5: Remove the highest edge(s) in (in case of same weights, edges must

be cut simultaneously) to get subtrees

6: for do

7:  if children of then

8:   All children nodes in this subtree are assigned as “noise”.

9:  else

10:   assign a new cluster to subtree

11:  end if

12: end for

13: until some stopping condition is satisfied.