Re: [igraph] Re: Question about community detection

[Lin Freeman]

Hi,

For other ways of evaluating any partitioning into communities, you might want to read:

Sailer, L. and S. Gaulin 1984 Proximity, Sociality, and Observation the Definition  of
   Social Groups. American Anthropologist 86: 91-98.

L. C. Freeman, S. C. Freeman and A. G. Michaelson  "How Humans See Social Groups: A Test of the Sailer-Gaulin Models." Journal of Quantitative Anthropology, 1, 1989, 229-238

As to whether a partition that yields a point as a community, consider a null graph containing N points and no edges.  Such a graph must display N communities.

Best,
Lin Freeman

On Mon, Dec 21, 2009 at 6:42 AM, Gábor Csárdi <address@hidden> wrote:

On Mon, Dec 21, 2009 at 3:01 AM, zhengjun chen <zhjchen.sa@gmail.com> wrote:
> Hi,

> I choose the community division with highest modularity value. What I find
> is that some communities just contain one node.
> So, I am wondering it is a good community division or not. So, is there any
> other way to demonstrate the community division is good or not?

I don't know of any. Theoretically it is possible to calculate the
best possible division in terms of modularity, but this is not
feasible for large graphs.

Btw. for me the other methods gave better modularity values on all the
graphs i have tried, so you might want to try them. They are also more
stable.

Gabor

> On Sun, Dec 20, 2009 at 10:40 AM, Gábor Csárdi <address@hidden> wrote:
>>
>> On Sun, Dec 20, 2009 at 4:26 PM, zhengjun chen <zhjchen.sa@gmail.com>
>> wrote:
>> > Hi,
>> > I am using Modularity to evaluate the goodness of community division.
>> > In order to get the largest modularity, a while loop is used to record
>> > the
>> > modularity value of different "step", which is a parameter of function
>> > "igraph_community_leading_eigenvector()". In such way, I record 100
>> > modularity value, and find the largest one corresponding to the "step"
>> > value
>> > 48. I initialized the membership vector and arpack option parameters for
>> > every loop.
>> > However, to verify the result, I set the "step" to 48, and rerun the
>> > program
>> > (just one time). The modularity value I get this time is different from
>> > that
>> > obtained by the while loop. I am a bit confused.
>>
>> Hi,
>>
>> one possible explanation is that the eigenvector solver is quite
>> unstable, and might result different splits. But it is fine to run it
>> a couple of times and then keep the version with the highest
>> modularity.
>>
>> Gabor
>>
>> > On Sat, Dec 19, 2009 at 5:13 PM, zhengjun chen <zhjchen.sa@gmail.com>
>> > wrote:
>> >>
>> >> Hi,
>> >> I have a question about community detection in
>> >> Igraph."igraph_community_leading_eigenvector()" function is used to
>> >> detect
>> >> communities.
>> >> My question is what is the meaning of "steps" in this function. Does it
>> >> mean the number of community split operation?
>> >> When I change this value from 10 to 100, more and more communities have
>> >> been detected. So, the parameter "steps" should be set to what?
>> >> In addition, I have no idea about how many communities actually exist.
>> >> So,
>> >> what can I do to verify that the communities detected is good enough?
>> >> I am also a bit confused of the structure igraph_arpack_options_t. Now
>> >> I
>> >> just initialize it to default values.
>> >>
>> >> --
>> >> Thanks
>> >> zhengjun
>> >>
>> >> Graduate research assistant
>> >> Dept of Computer Science and Engineering
>> >> Lehigh University
>> >
>> >
>> >
>> > --
>> > Thanks
>> > zhengjun
>> >
>> > Graduate research assistant
>> > Dept of Computer Science and Engineering
>> > Lehigh University
>> >
>> > _______________________________________________
>> > igraph-help mailing list
>> > address@hidden
>> > http://lists.nongnu.org/mailman/listinfo/igraph-help
>> >
>> >
>>
>>
>>
>> --
>> Gabor Csardi <address@hidden>     UNIL DGM
>>
>>
>> _______________________________________________
>> igraph-help mailing list
>> address@hidden
>> http://lists.nongnu.org/mailman/listinfo/igraph-help
>
>
>
> --
> Thanks
> zhengjun
>
> Graduate research assistant
> Dept of Computer Science and Engineering
> Lehigh University
>
> _______________________________________________
> igraph-help mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/igraph-help
>
>



--
Gabor Csardi <address@hidden>     UNIL DGM


_______________________________________________
igraph-help mailing list
address@hidden
http://lists.nongnu.org/mailman/listinfo/igraph-help

--

This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.

-- 
Linton C. Freeman
Institute for Mathematical Behavioral Sciences and
Department of Sociology
School of Social Sciences  SSPA 2143
University of California
Irvine, CA 92697-5100

Office:             (949) 824-6698
Home (CA):      (949) 494-6139
          (FL):      (941) 778-1074
Secretary:        (949) 824-3663
FAX:                (949) 824-3733