[Networkit] NetworKit - Community detection

Christian Staudt christian.staudt at kit.edu
Thu Jan 15 22:51:27 CET 2015


Hi Israa,

> Based on your experience in network analysis, what other network properties I can examine when detecting communities?

What does your graph represent? Are there node attributes that can be used to interpret and validate the detected communities? What is the role of hub nodes with respect to communities? Which nodes are central or peripheral in a community? Which communities are strongly connected to each other? How do communities look at different resolutions (play with the resolution parameter of PLM)?

> and is it possible to visualize the communities given that I have a graph of 50M nodes and 300M edges?

That’s not easy. Even if you draw a coarsened version of the graph in which a community is a node, it’s relatively big. What are you hoping to understand from a drawing?

Looking forward to having a look at your thesis.
Chris

Am 15.01.2015 um 15:48 schrieb Isra Al Qasem <isra_aq at hotmail.com>:

> Hello Chris,
> 
> Yes I tried both, I looked at:
> 
> the total number of evolving communities over time
> the size of the largest and avg community over time
> modularity values
> 
> Based on your experience in network analysis, what other network properties I can examine when detecting communities? and is it possible to visualize the communities given that I have a graph of 50M nodes and 300M edges?
> 
> Thanks for your prompt replies. I really appreciate your help. 
> Hopefully this semester I will finish my thesis and will send you a copy:)
> 
> Regards,
> Israa
> 
> 
> 
> From: christian.staudt at kit.edu
> Subject: Re: NetworKit - Community detection
> Date: Thu, 15 Jan 2015 14:52:14 +0100
> To: isra_aq at hotmail.com
> 
> Hi Israa,
> 
> NetworKit provides mainly two modularity-maximizing algorithms for partitioning the node set into disjoint communities: PLP and PLM. See this paper for details: 			http://parco.iti.kit.edu/staudt/attachments/publications/TPDS.pdf
> 
> I’m not entirely sure what you mean by characteristics - can you elaborate?
> 
> Chris
> 
> Am 15.01.2015 um 14:47 schrieb Isra Al Qasem <isra_aq at hotmail.com>:
> 
> Dear Chris,
> 
> Hope this email finds you well.
> 
> I'm analyzing the evolution of communities in a large graph over time and I  want to know
> what kind of characteristics I can examine via the community detection algorithms offered in NetworKit?
> 
> Best Regards,
> Israa

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ira.uni-karlsruhe.de/mailman/private/networkit/attachments/20150115/7d5598f5/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 495 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <https://lists.ira.uni-karlsruhe.de/mailman/private/networkit/attachments/20150115/7d5598f5/attachment.sig>


More information about the NetworKit mailing list