Graph with Weighted Nodes

I have file with 3 columns N1 N2 W nodes labels and weight. My goal to print a graph with outgoing nodes with W greater than 200 and corresponding to them input nodes with W greater then 20. What I tried, filter only lower number for W or not working.

Hi @izaychik63 -

Can you please post an example workflow of what you’ve tried already, and what you’d expect the output to look like?

I have WF below

On the output I’d like to have oriented graph with nodes size corresponding to the outgoing edges weight pointing to the nodes with incoming edges with the size corresponding to the incoming weight.
The idea to have 2 type of association with 2 reports with big nodes on the both sides of the the oriented graph.

  1. Identify top providers referring out of the company.
  2. Identify top referrals as binary relations (prov. A to prov. B has > 20 referrals).

@ScottF, is it possible to find a cliques with edge weight bigger than say 20. As a result, I’d like to have a graph picture and a table with pares Out_Node, In_Node.

Hi @izaychik63 -

Sorry for the delay - I’m still playing catchup after the Fall Summit. This isn’t really my area of expertise, so let me see if I can find another KNIMEr to weigh in.

1 Like

Hi izaychik63,
I don’t think there is a straight solution to do this. But you can use the standard KNIME clustering nodes e.g. Hierarchical Clustering on the network adjacency matrix as show here:

If you are interested in a biclique you can also use the Item Set Finder (Borgelt) node together with the Subset Matcher node as shown in this example workflow. Since the Subset Matcher also allows for a number of mismatches you can also find quasi bicliques where some connections are missing between the two partitions.

Once you have detected the subgraph you need to extract the edge weight and group over the whole table to get the sum of all edge weights per subgraph. Take a look at the DrugBank workflow for an example on how to extract the subgraph after the biclique detection.


Thank you, @tobias.koetter. It looks pretty comprehensive.

This topic was automatically closed 182 days after the last reply. New replies are no longer allowed.