Hi Philipp,
these SNA (Social Network Analysis) were so much the hype, don't
really know if they shine so much
these days.
This is a work from the _physics_ point of view, as an attempt to
observe natural laws (statistical physics, complex networks).
In this sense, this poses a somewhat interesting
question, as to how do or do not our interactions yield natural structures.
These "natural structures" are found in gene, food, airport, bone
cavities, sexual etc networks.
In what way are these (email) interaction
networks the same or different from these other networks, if in any sense?
Are there good uses of these structures that the community can make?
In a first attempt to characterize these interaction networks' topology,
reported on the article, there were three main results:
1) stability in which criteria (measures) give better resolution for
understanding (classifying) activity of these interaction networks
(seen in PCA composition by original measures). Connectivity
is mandatory, followed by asymmetries of participation and relations,
and, in third place, stands clustering as formation of community
structures ("all knows all").
2) There is clear deviation from an uniform distribution of
interaction. This deviation
reveals 3 sectors: periphery (~80% participants), intermediary (15%
participants)
and hub sectors (~5% participants). Exact method for this division
is detailed in text. Last pages of the article:
http://sourceforge.net/p/labmacambira/fimDoMundo/ci/master/tree/textos/evolutionSN/evsn.pdf?format=raw
has figures that exhibit this ternary division as networks evolves,
within a fixed number of messages.
3) Stability of activity along time, with respect to seconds of a
minute, minutes of an hour, days of the month. Concentration
of activity along hours of the day, days of the week and months on the year.
X) These results hold for all lists analysed, including all 4 lists
selected for formal report.
Thanks for pointing "big data", I did not mean to go that way, but for
extracting information all my 8GB RAM is used, being
selective at each run and running tens of times. ( These scripts:
sourceforge.net/p/labmacambira/fimDoMundo/ci/master/tree/python/toolkitGMANE/ )
Regards,
Renato
2013/12/22 Philipp Überbacher <murks@xxxxxxxxxxxxx>:
> On Sun, 22 Dec 2013 20:12:53 -0200
> Renato Fabbri <renato.fabbri@xxxxxxxxx> wrote:
>
>> Dear LAU,
>>
>> :::
>>
>> ---------- Forwarded message ----------
>> From: Renato Fabbri <renato.fabbri@xxxxxxxxx>
>> Date: 2013/12/20
>> Subject: GMANE and complex networks
>> To: linux-audio-dev@xxxxxxxxxxxxxxxxxxxx
>>
>>
>> Dear LAD,
>>
>> In studying complex network (a doctorate research),
>> I got into interaction networks because of its utility for
>> understanding social systems.
>>
>> This lead me to GMANE database:
>> gmane.org
>> in which LAD, LAU, LAA (i think), and about 20 thousand other lists
>> are hosted as public and with data available via RSS.
>>
>> After experimentations with some lists, in writing results in an
>> article format, I chose 4 lists: the GNU C++ stdlib development list
>> (official perhaps), LAU, LAD and Metareciclagem, a
>> gadget-media-activist list from Brazil.
>> This article was sent to arXiv:
>> arxiv.org/abs/1310.7769
>> and is currently being revised by authors, with latest version here:
>> http://sourceforge.net/p/labmacambira/fimDoMundo/ci/master/tree/textos/evolutionSN/evsn.pdf?format=raw
>> Some visualizations of these networks in evolution are in:
>> http://www.youtube.com/watch?v=-t5jxQ8cKxM&list=PLf_EtaMqu3jU-1j4jiIUiyMqyVSzIYeh6
>> and:
>> http://hera.ethymos.com.br:1080/redes/python/autoRede/escolheRedes.php
>>
>> Among all options available for doing this research, I chose LAD and
>> LAU with esteem. This lists were quite helpful to me in many
>> occasions, specially in the period 2005-2009. Anyway, this raises a
>> question about this kind of analysis, if it is desirable, invasive in
>> public lists/data. As they are publicly accessible, users should have
>> access also to what kind of information one is able to extract from
>> such data? Or should it be restricted to enterprises, government
>> parties and individuals not sharing about it? I number participants,
>> so names don't appear on results and even in the process of data
>> mining, but should that be? Should that hold for public data?
>> Of course, this discussion might make sense only when there are no
>> aggressive intents, such as developing interfaces to expose someone,
>> which is probably not cool in any case.
>>
>> Cheers!
>> //r
>>
>>
>> --
>> GNU/Linux User #479299
>> labmacambira.sf.net
>
> Hi Renato,
> I skimmed over the results and wonder what the purpose of this exercise
> was (besides the degree). What do the results tell you?
>
> I know that 'big data' and network analysis is all the hype, but I fail
> to see anything interesting there.
>
> Regards,
> Philipp
> _______________________________________________
> Linux-audio-user mailing list
> Linux-audio-user@xxxxxxxxxxxxxxxxxxxx
> http://lists.linuxaudio.org/listinfo/linux-audio-user
--
GNU/Linux User #479299
labmacambira.sf.net
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user@xxxxxxxxxxxxxxxxxxxx
http://lists.linuxaudio.org/listinfo/linux-audio-user
