Adjacency matrix clustering using spectral (cluster) network in Python

Question

I have the following dataset:

firm_id firm_id_
1         2
1         4
1         5
2         1
2         3
3         2
3         6
4         1
4         5
4         6
5         4
5         7
6         3
...

This data says for exampe that firm_id = 1 is directly connected to firm_id = 2, 4, and 5 and indirectly connected (within two paths) to firm_id = 3, 6, and 7. I can use some Python package like networkx to build the network of firm's connectivity. Now, I want to use Spectral Clustering (I guess this the correct methodology) to form clusters based on distance (number of edges separating each firm) and see how these clusters are connected to each other.

I would first define an adjacency matrix W of the above data. I then use enter image description here where dist is the distance between firm i and firm j, and c is a scale parameter to each element in W and then compute the Laplacian matrix (see here for example).

Now my question is can Spectral Clustering give me the link between each clusters and how far apart are the clusters (how many edges separate the clusters)? I was thinking to use this, the scikit package in Python but I have no idea how I can generate the links between clusters using sklearn.cluster.

Plug4 · Accepted Answer · 2014-04-17 12:38:10Z

2

Community detection network is what I needed:

http://perso.crans.org/aynaud/communities/

answered Apr 17, 2014 at 12:38

Plug4

3,97811 gold badges55 silver badges87 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

Has QUIT--Anony-Mousse · Accepted Answer · 2014-04-18 09:02:55Z

1

For spectral clustering and these approaches to work well, you need to have similarities.

Your data seems to be solely a graph, i.e. edges that connect instances. You should be looking at graph partitioning and maybe community detection algorithms that work solely on the graph structure, and do not assume you have a continuous measure of similarity.

answered Apr 18, 2014 at 9:02

Has QUIT--Anony-Mousse

77.8k14 gold badges145 silver badges198 bronze badges

2 Comments

Plug4 Over a year ago

thanks for the reply. You are right. I don't have similarities. I only have linkages (where weight is always 1 between nodes that are connected). I decided to use this algo perso.crans.org/aynaud/communities in python and I can form communities but I have no idea how the communities are connected. Would you have a suggestion to figure out how the communities would be connected?

Has QUIT--Anony-Mousse Over a year ago

I have no experience on this kind of data. All I've seen on social network analysis I found pretty much random, obvious or otherwise uninteresting.

Collectives™ on Stack Overflow

Adjacency matrix clustering using spectral (cluster) network in Python

2 Answers 2

Comments

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related