Hello,

I am dealing with three relation sets of the same actors I will explain ahead. These networks were built through Visone. Considering “Knime Network Plugin” resources/tools, I am asking for advice about which could be applied to improve this network analysis (actually, I have one objective question, but further advice will be welcome).

Follow attached the image of these networks; the different node size is related to the real size of the actors which are corporations.

For each set I measured (1) density and (2) correlation between degree and size:

a) Competition Network: the actors mentioned who are their competitors.

Density: 0.48

Correlation degree x size: 0.54 (p-value = 0.01)

b) Acting Network: the actors mentioned they know each other and exchange some general resources or information.

Density: 0.40

Correlation degree x size: 0.76 (p-value = 5.2)

c) Cooperation Network: the actors mentioned they effectively cooperate with each other in a “real” network structure.

Density: 0.19

Correlation degree x size: 0.45 (p-value = 0.03)

Considering correlation “degree x size”, it seems that in competition and cooperation networks, with moderate intensity, the bigger the actor, the bigger the likelihood of the competition and cooperation behaviour.

Considering density, it’s is clear that relations of competition are bigger than the others. Besides, it is suggestive that much of the cooperation network relations exist inside the other networks. I could see (not actually calculate) that even being competitors, same actors “act” or “cooperate” with each other (e.g. there are relations between actors B and T along the three networks; great part of actor relations in cooperation network exist inside competition network). So, is there any way to calculate the amount of these same actors who relations are (remain the same) in the different networks? I would like to find something like 80% of cooperative relations exist inside competition network.

As I told before, besides this question, any further advice/suggestion of analysis using Knime Network Plugin will be welcome!

Hello,

I could identify more exactly what I need from Knime!

Let me simplify the idea of the networks I send image attached before.

Cooperation Network

id            id

a             b

a             c

c              e

Competition Network

id            id

a             b

a             c

e             g

f              d

So, seeing the example above, the following actor pairs is matching:

id            id

a             b

a             c

I would like to proceed with this matching. May anyone tell me if it is possible and how to do? Also, I need to begin from a “viz input connector” once I will load a network from Visone.

Ps.: Originally the networks was made through Visone from matrix data, not exactly link lists as in my example. But once I import it to Knime, could I extract the link list and proceed with the matching considering link list data structure?

Well, I dont know much about Network Pairings or the network plugin, but the example you mention of wanting to pull back matching networks should be easily possible.

First use the "Column Combiner" node in Data Manipulation/Column/Split&Combine to combine together the two ID columns in "Cooperation Network", and then do the same for the "Competition Network".

Now take the "Reference Row Filter" node in Data Manipulation/Row/Filter, and input both the "Cooperation Network" and "Competition Network" flows into the node. In the configuration of the node, choose "Include Rows From Reference Table", and in the drop downs, choose the new columns you created which has the combined IDs. The output will then give as you showed in the example, the matching actor pairs.

I hope this is of help,

Simon.

Hi Simon,

Thanks for your answer! I am looking forward to try it. But first I need to know if Knime is able to extract/obtain the link list (the format I mentioned before "id id") from a “viz input connector” (network minning plugin). Does anyone know if it is possible and how to?

Best,

the "Edge Table" creates an edge list that includes the edge id and label as well as the label and id of the incident nodes e.g.

Edge id Edge label Node id Node label
e1 edge1 n1 node1
e1 edge1 n2 node2

Bye,

Tobias

Hi Tobias,

Thanks for information. I am handling this now.

Bye,