Network Mining - how to extract clusters?

KNIME Extensions
1

Hi,

I'm playing around with the network mining nodes and I have a network which consists of three unconnected clusters (see picture). How can I extract these cluster with their members?

If there would be a single edge between such clusters, how could I remove them?

2

Hi,
you can use the "Network To Row" node to split-up a network into its unconnected components. The result is a data table that contains for each unconnected component a row with the component and its members. You can than further process each component when looping over the data table as demonstrated in the example workflow 01202005_networkLooping (see the example page).
Currently we do not have implemented methods for graph clustering such as cut based approaches but we will introduce some in one of the next releases. However you can delete the edge manually before extracting the unconnected components using the object id filter.
Bye,
Tobias
 

1 Like
3

Thanks a lot!

These tips are very useful.

4

Hello, I’m looking for the same solution. Is already implemented in KNIME? I want to find the memberships of the clusters.
I had found an R implementation, but I would like to do it in KNIME.

The R code is:

library(igraph)

t1 ← t0 %>%
filter(ISO2 == country) %>%
graph_from_data_frame(directed = FALSE) %>%
clusters(mode=“strong”)

t1$membership

From:
https://stackoverflow.com/questions/30105232/visualize-strongly-connected-components-in-r