How to get non-singleton cluster ids in scipy hierachical clustering

Question

According to this we can get labels for non-singleton clusters.

I tried this with a simple example.

import numpy as np
import scipy.cluster.hierarchy         


        

Answer 1

I think the document is not very clear at this part and the sample code in it is not even operational. But it is clear that 1 means the 2nd observation and (3) means there are 3 observation in that node.

If you want to know what are the 3 obs. in the 2nd node, if that is your question:

In [51]:
D4=dendrogram(linkage_matrix,
              color_threshold=1,
              p=4,
              truncate_mode='lastp',
              distance_sort='ascending')
D7=dendrogram(linkage_matrix,
              color_list=['g',]*7,
              p=7,
              truncate_mode='lastp',
              distance_sort='ascending', no_plot=True)  
from itertools import groupby
[list(group) for key, group in groupby(D7['ivl'],lambda x: x in D4['ivl'])]
Out[51]:
[['1'], ['6', '0', '3'], ['2'], ['4', '5']]

The 2nd node contains obs. 7th, 1th and 4th, and the 2th node contains the 5th and the 6th observations.