Bitcoin Forum
August 20, 2019, 04:50:44 AM *
News: Latest Bitcoin Core release: 0.18.0 [Torrent] (New!)
 
   Home   Help Search Login Register More  
Pages: [1]
  Print  
Author Topic: Global structure of merit networks  (Read 538 times)
sncc
Hero Member
*****
Offline Offline

Activity: 532
Merit: 509


View Profile
May 08, 2018, 02:45:16 PM
Last edit: May 08, 2018, 04:07:43 PM by sncc
Merited by theymos_away (25), malevolent (10), Foxpup (9), TMAN (6), LoyceV (5), DarkStar_ (5), Lutpin (5), zentdex (4), mprep (3), EFS (3), botany (3), DdmrDdmr (2), OgNasty (1), coinlocket$ (1), bitmover (1), 1miau (1), theyoungmillionaire (1)
 #1

I have been interested in applying some tools for social network analysis visualization to merit transactions, and finally I have had some time to analyze global structure of merit networks.  

"gif movie"

In the gif movie above and figures below, forum accounts and their merit transactions are expressed by nodes and edges, respectively.  In the above gif movie, time evolution of every nodes are governed by attractive and repulsive force which are calculated from the number of total sent merits between accounts by using ForceAtlas2 [1].  The nodes are classified and given different colors depending on the communities they belong to, which are identified by using the Louvain method [2].  These methods visualize the strength of merit connections explicitly, and detects its substructure algorithmically.

After waiting for a while for stabilization of the system, we obtain the global structure of merit networks:

"Fig. 1"

We see there are several communities in the forum.  Accounts in the same communities have close relationship through merit transactions.  Let us take a closer look of each community.

Purple (theymos, suchmoon, Vod,...): mixture of English boards such as Bitcoin, Altcoin, Meta, Beginners & Help etc. and small local boards
Green (JayJuanGee, LastoftheV8s, vapourminer,...): Economy

"Fig. 2"

Purple and green groups are the largest and closely connected groups.  While purple looks like a mixture of several sections, somehow Economy section is deteceted as an independent goup, as there are certain people who focus on Economy section.  Small local boards such as Japanese board are included in the purple group.

Other groups are top 3 largest local boards.

Black (dbshck, pandukelana2712,...): Local > Bahasa Indonesia (Indonesian)

"Fig. 3"

Orange (EFS, AlyattesLydia,...): Local > Türkçe (Turkish)

"Fig. 4"

Green (Ranyar, RuSS512,...): Local > Pyccкий (Russian)
Red (Co1n, poptop,...) : Local > Pyccкий (Russian)

"Fig. 5"

Green and Red are both Russian community implying that it is also sufficiently large to exhibit its substructure.  From the previous analysis by zentdex [3], the top 3 merit distributed local boards are Russian, Turkish, Indonesian local boards.  The above result is thus consistent.  Furthermore, the above analysis sheds light on their relative distance from other boards:  Among the three local boards, Russian community is the largest but at the same time the most far from the English boards.  On the contrary, Indonesian local board is the smallest among the three, but nearest to the English boards, implying that they are sending merits each other.

We note that there are several large nodes that do not belong to a community but has different colors from their vicinity: paxmao, Ognasty, ui_zakharchenko, explorder, xandry fall into such a class.  Why they are independent from any group?

The interpretation is that their merit histories are peculiar and different from others.  paxmao and Ognasty look like sending merits to a lot of sections, which is why the algorithm could not tell which communities they are belonging to.  It implies that they have been trying hard to distribute merits to the entire forum.  

On the other hand, explorder and ui_zakharchenko are the OPs of the explanations of merit system [4] and famous bounty managers [5], which are very popular in Russian local board.  Since their merit transactions are mainly within those threads, they respectively form independent groups.  

Finally, xandry has sent most of smerits to "(Deleted/Off-limits/Ignored)" so I assumed his/her transactions also mainly remain within a single topic, but it is still not clear to me how to interpret his/her merit history:

"Fig. 6"

I assumed he/she is focusing on some particular thread and evaluating users posted there, but it is not clear to me why he/she needs to send merits so frequently.  Maybe some very active thread was deleted recently together with a lot of posts which were merited?  In any case it is indeed a special case, which is automatically revealed by the above algorithm.

In these cases, they are recognized as small independent communities and have different colors.  However, they are moderately belonging to English and Russian communities which can be read off from their distance to those communities.

Of course, a related analysis is the wonderful analysis and implementation by DdmrDdmr [3], which allows us to check personal merit networks very easily.  The thread appeared when I was learning and preparing this analysis, and I was actually afraid that nothing was left for further work after his/her analysis.  Fortunately the above analysis for global structure seems complementary to his/her work.  Now we know personal and global structures of merit networks.  



References:
[1] M. Jacomy , T. Venturini, S. Heymann, M. Bastian, "ForceAtlas2, a Continuous Graph Layout Algorithm for Handy Network Visualization Designed for the Gephi Software", PloS one 9 (6), e98679 (2014).
[2] V. Blondel, J. Guillaume, R. Lambiotte, E. Mech, "Fast unfolding of communities in large networks", J. Stat. Mech. P10008 (2008), [arXiv:0803.0476] (http://findcommunities.googlepages.com).
[3] zentdex, "Where the Merit Pours?", Bitcointalk, 3093768, March 09, (2018).
[4] explorder, "Merit - чтo этo и кaк этим пoльзoвaтьcя", Bitcointalk, 2818398, January 24 (2018).
[5] ui_zakharchenko, "Пoлeзнocти для бayнтиcтoв", Bitcointalk, 2845483, January 29, (2018).
[6] DdmrDdmr, "Our very own sMerit Network Picture", Bitcointalk, 3395255, April 25 (2018).
Advertised sites are not endorsed by the Bitcoin Forum. They may be unsafe, untrustworthy, or illegal in your jurisdiction. Advertise here.
1566276644
Hero Member
*
Offline Offline

Posts: 1566276644

View Profile Personal Message (Offline)

Ignore
1566276644
Reply with quote  #2

1566276644
Report to moderator
1566276644
Hero Member
*
Offline Offline

Posts: 1566276644

View Profile Personal Message (Offline)

Ignore
1566276644
Reply with quote  #2

1566276644
Report to moderator
lobcmt2
Full Member
***
Offline Offline

Activity: 420
Merit: 154



View Profile
May 08, 2018, 04:04:45 PM
 #2

Thank you sncc for such extremely high-quality, informative, and helpful topic.
Would you mind sharing the data which you used for your analysis, please.
I would like to do some statistical analysis by myself.
Honestly, I don't want to struggle with you to earn merits, lol.  Grin
DdmrDdmr
Hero Member
*****
Offline Offline

Activity: 588
Merit: 2569

There are lies, damned lies and statistics. MTwain


View Profile WWW
May 08, 2018, 04:29:15 PM
 #3

<...>

Nice work there sncc. It does seem complementary to what I did before, hilighting certaing global communities of connected groups within a rather large network in a nice visual manner.

I also did a similar network on trust, although the data in the trust.txt file is rather limited, but it enables you to retrieve who trusts you and who you are trusted by ( A picture of what our Trust/Distrust Network looks like).

sncc
Hero Member
*****
Offline Offline

Activity: 532
Merit: 509


View Profile
May 08, 2018, 10:15:13 PM
 #4

Thank you sncc for such extremely high-quality, informative, and helpful topic.
Would you mind sharing the data which you used for your analysis, please.
I would like to do some statistical analysis by myself.
Honestly, I don't want to struggle with you to earn merits, lol.  Grin
I just used the theymos' data uploaded at
https://bitcointalk.org/merit.txt.xz
All merit transactions during the last 120 days are listed there.

Nice work there sncc. It does seem complementary to what I did before, hilighting certaing global communities of connected groups within a rather large network in a nice visual manner.

I also did a similar network on trust, although the data in the trust.txt file is rather limited, but it enables you to retrieve who trusts you and who you are trusted by ( A picture of what our Trust/Distrust Network looks like).
Thanks, yes I am aware of your another work, and planning to do the same game for the trust as well after studying merit networks a little bit more.
OgNasty
Donator
Legendary
*
Offline Offline

Activity: 3010
Merit: 1705


I 💚 Bitcoin


View Profile
May 08, 2018, 10:26:33 PM
 #5

We note that there are several large nodes that do not belong to a community but has different colors from their vicinity: paxmao, Ognasty, ui_zakharchenko, explorder, xandry fall into such a class.  Why they are independent from any group?

The interpretation is that their merit histories are peculiar and different from others.  paxmao and Ognasty look like sending merits to a lot of sections, which is why the algorithm could not tell which communities they are belonging to.  It implies that they have been trying hard to distribute merits to the entire forum.  

Very cool.  I have been going out of my way to hunt across many sections for merit worthy posts to make sure that merits are being distributed as intended.  It is nice to see that I've done such a good job the algorithm could not even classify me.  It is also nice to be recognized as having been, "trying hard to distribute merits to the entire forum."

Thanks.  Kudos for a cool use of available data.

sncc
Hero Member
*****
Offline Offline

Activity: 532
Merit: 509


View Profile
May 09, 2018, 01:28:42 PM
 #6

We note that there are several large nodes that do not belong to a community but has different colors from their vicinity: paxmao, Ognasty, ui_zakharchenko, explorder, xandry fall into such a class.  Why they are independent from any group?

The interpretation is that their merit histories are peculiar and different from others.  paxmao and Ognasty look like sending merits to a lot of sections, which is why the algorithm could not tell which communities they are belonging to.  It implies that they have been trying hard to distribute merits to the entire forum.  

Very cool.  I have been going out of my way to hunt across many sections for merit worthy posts to make sure that merits are being distributed as intended.  It is nice to see that I've done such a good job the algorithm could not even classify me.  It is also nice to be recognized as having been, "trying hard to distribute merits to the entire forum."

Thanks.  Kudos for a cool use of available data.
Thanks, yes it is very clear from the analysis that you are an outstanding merit sender.  Congrats for your efforts beating the algorithm.
Lauda
GrumpyKitty
Legendary
*
Offline Offline

Activity: 2310
Merit: 2102


Modern Liberalism is a Mental Disorder


View Profile
May 09, 2018, 02:58:54 PM
Merited by paxmao (1)
 #7

You should provide a way for others to explore this visualization (website?). I can already see the main sources of account farming for some sections. Surprise, surprise. Roll Eyes

████████████████████████████
████████▀▀ █▀ █▀ ▀██████████
█████████▄ ▄▄▄▄▄▄███████████
██████████▀     ▀  ▀████████
███████▀ ▀  ▄█▀▀▀█▀▀████████
██████▄      █▄  ▀▀  ▀██████
██████         ▄▄█▄ ▄ ▀█████
█████ ▄         ▀▀ ▄ ▀ █████
██████▌          █▀█▀ ▐█████
███████  ▄▌         ▄ ██████
████████▄█         ▄████████
█████████▀     ▄▄ ▄█████████
████████████████████████████
.JACKMATE'S...........
.
MAJESTIC..
████████████████████████
███████████████████████
████████████████████████
████████████████████████
████████████████████████
████████████████████████
████████████████████████
████████████████████████
████████████████████████
████████████████████████
████████████████████████
████████████████████████
████████████████████████
.
..WIN 1 BITCOIN ON EVERY PREMIER LEAGUE MATCHDAY..
████████████████████████████████
████████████▀█▀ ▀█▀█▀███████████
███████████▄ ▄▄▄▄▄▄▄████████████
███████████▀▀▄▄▄▄▄▄▄▄███████████
█████████▀▄ ██▀▄▄▄ ▀ ▄▀█████████
███████▀ ▀█████▄▄▄█▄▄▄██████████
███████▀▄████████▀  ▀█ █▐███████
███████ ▀█████████▄█▀▀██ ███████
████████ ███▀██████ ▄ ██ ███████
████████▌▐▀▄ ██████████ ▄███████
█████████▄██▌▐█████▀██ █████████
████████████▄▀▀▀▀▀▄ ▀▄██████████
████████████████████████████████
.
.JOIN US - IT'S FREE! .
bitmover
Hero Member
*****
Offline Offline

Activity: 574
Merit: 970



View Profile
May 09, 2018, 07:00:15 PM
 #8

I was very happy that with some zoom I could find my name and even found my local board .

You should provide a way for others to explore this visualization (website?). I can already see the main sources of account farming for some sections. Surprise, surprise. Roll Eyes

that would be a great idea. Where we could make zoom in to see these connections more clearly.

sncc
Hero Member
*****
Offline Offline

Activity: 532
Merit: 509


View Profile
May 09, 2018, 09:39:49 PM
Last edit: May 14, 2018, 09:01:19 PM by sncc
 #9

I was very happy that with some zoom I could find my name and even found my local board .

You should provide a way for others to explore this visualization (website?). I can already see the main sources of account farming for some sections. Surprise, surprise. Roll Eyes

that would be a great idea. Where we could make zoom in to see these connections more clearly.
Not website, the movie and figures are generated by using igraph on python and Gephi (in Ref. [1]) with merit data provided by theymos.  There are many softwares in this field, so you may find your favorite ones.  

As you guys noted, the network analysis is a powerful tool to identify suspicious merit transactions.  Now I am focusing on small scale structure and mutual merit transactions.  Will report the follow up analysis soon*.



* "Merit network analysis: merit rank distribution and satellites"  
https://bitcointalk.org/index.php?topic=3759121.0
zentdex
Full Member
***
Offline Offline

Activity: 196
Merit: 274


View Profile
May 17, 2018, 01:00:34 PM
 #10

Magnificent!
I can't imagine how you able to create something like that.
Is it possible to add time stamps per frame...
sncc
Hero Member
*****
Offline Offline

Activity: 532
Merit: 509


View Profile
May 20, 2018, 10:41:15 AM
 #11

Magnificent!
I can't imagine how you able to create something like that.
Is it possible to add time stamps per frame...
The gif movie shows how the method detects the structure, and does not reflect how actual merit networks are constructed in time.  So there is no time stamp here.

Can you add Philippines Local board too? I posted there often. thanks
The subcommunities are automatically identified and colored by the method and I only interpreted them as local boards or subsections.
Pages: [1]
  Print  
 
Jump to:  

Sponsored by , a Bitcoin-accepting VPN.
Powered by MySQL Powered by PHP Powered by SMF 1.1.19 | SMF © 2006-2009, Simple Machines Valid XHTML 1.0! Valid CSS!