SNA Analysis

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

SNA Analysis

NetLogo-Users mailing list
 blockquote, div.yahoo_quoted { margin-left: 0 !important; border-left:1px #715FFA solid !important; padding-left:1ex !important; background-color:white !important; } Hi All, Not sure if my question is relevant to this group or not. But I hope I will get some help here. 
I'm visualising millions of nodes in social network .. Bank customers based on their shared interests, joint accounts, transactions etc. 

Is there any clustering analysis tools or techniques/algorithm you may would suggest that I could apply to narrow down from large network to small network ? 

An obvious goal here is to find communities to recommend or do any predictive analysis.
Thanks ,  

Sent from Yahoo Mail for iPhone
Reply | Threaded
Open this post in threaded view
|

Re: SNA Analysis

NetLogo-Users mailing list
Hi,

Since this is a mailing list for netlogo I will start with a netlogo related answer.
The best way to identify communities in NetLogo is through the modularity maximisation method in the nw extension.
The method is called the Louvain method: nw:louvain-communities. <> The basic idea of the method is the minimisation of the number of links between communities while maximising the number of links inside communities. The advantage of this method is that you will not need to decide how many communities you want, the algorithm decides for you. For details see https://arxiv.org/abs/0803.0476 <https://arxiv.org/abs/0803.0476>.

If you are looking for tools able to analyze your network outside of NetLogo, Gephi (https://gephi.org) uses the same algorithm for cluster identification and allows for the visualisation of the communities by colouring the nodes. Gephi also includes some basic SNA metrics (eigenvector, betweenness, page rank, clustering and so forth) and is capable of visualising millions of nodes (with a computer up to the task).

If you are more interested in computing metrics (since visualising millions of nodes might result in a huge hairball) R (or python) might be your best bet. The ‘igraph’ package, available in both languages, has everything you might need (http://igraph.org/redirect.html). You can chose not to visualise the network which will allow you to speed up computations.

Pajek has a special version, Pajek XXL for very large networks, but i believe it offers less diverse metrics compared to R (http://vlado.fmf.uni-lj.si/pub/networks/pajek/).

Cytoscape is one of the most user friendly tools (http://www.cytoscape.org), but i’m not sure if it can handle such a large network. In addition, Cytoscape was developed for biological networks and therefore the cluster identification methods are also tuned for biological networks.

All the tools described above are free. There must be many more but these are the ones I am familiar with.

Hope this helps


==============================================
Johannes van der Pol
Post-doc, plateforme VIA-INNO, GREThA UMR-CNRS 5113
PhD consultant Labex TRAIL
University of Bordeaux
Avenue Léon Duguit, 33600 Pessac
Office F346a





> Le 29 juin 2017 à 00:03, Bandeh Ali Talpur [hidden email] [netlogo-users] <[hidden email]> a écrit :
>
> Hi All,
>
> Not sure if my question is relevant to this group or not. But I hope I will get some help here.
>
> I'm visualising millions of nodes in social network .. Bank customers based on their shared interests, joint accounts, transactions etc.
>
> Is there any clustering analysis tools or techniques/algorithm you may would suggest that I could apply to narrow down from large network to small network ?
>
> An obvious goal here is to find communities to recommend or do any predictive analysis.
>
> Thanks ,
>
> Sent from Yahoo Mail for iPhone <https://yho.com/footer0>
>
>

Reply | Threaded
Open this post in threaded view
|

Re: SNA Analysis

NetLogo-Users mailing list
To add to what Johannes said, to do data science with networks I recommend Tulip. Also free and open, scriptable in Python, and very actively maintained (by Johannes's home university, Bordeaux U).

 http://tulip.labri.fr/TulipDrupal/ http://tulip.labri.fr/TulipDrupal/

 

 I have used both Pajek and Gephi, and in my opinion Tulip is far, far superior.
 

 There is now also a network package for Stata, if you are a Stata guy.
 

Reply | Threaded
Open this post in threaded view
|

Re: SNA Analysis

NetLogo-Users mailing list
In reply to this post by NetLogo-Users mailing list
 blockquote, div.yahoo_quoted { margin-left: 0 !important; border-left:1px #715FFA solid !important; padding-left:1ex !important; background-color:white !important; } This is really helpful Johannes. 
Thanks million. 
Ali 




Sent from Yahoo Mail for iPhone




On Friday, June 30, 2017, 5:44 p.m., Janpieter van der pol [hidden email] [netlogo-users] <[hidden email]> wrote:


 
Hi,


Since this is a mailing list for netlogo I will start with a netlogo related answer.The best way to identify communities in NetLogo is through the modularity maximisation method in the nw extension. The method is called the Louvain method: nw:louvain-communities. The basic idea of the method is the minimisation of the number of links between communities while maximising the number of links inside communities. The advantage of this method is that you will not need to decide how many communities you want, the algorithm decides for you. For details see https://arxiv.org/abs/0803.0476.
If you are looking for tools able to analyze your network outside of NetLogo, Gephi (https://gephi.org) uses the same algorithm for cluster identification and allows for the visualisation of the communities by colouring the nodes. Gephi also includes some basic SNA metrics (eigenvector, betweenness, page rank, clustering and so forth) and is capable of visualising millions of nodes (with a computer up to the task).
If you are more interested in computing metrics (since visualising millions of nodes might result in a huge hairball) R (or python) might be your best bet. The ‘igraph’ package, available in both languages, has everything you might need (http://igraph.org/redirect.html). You can chose not to visualise the network which will allow you to speed up computations.
Pajek has a special version, Pajek XXL for very large networks, but i believe it offers less diverse metrics compared to R (http://vlado.fmf.uni-lj.si/pub/networks/pajek/).
Cytoscape is one of the most user friendly tools (http://www.cytoscape.org), but i’m not sure if it can handle such a large network. In addition, Cytoscape was developed for biological networks and therefore the cluster identification methods are also tuned for biological networks. 
All the tools described above are free. There must be many more but these are the ones I am familiar with. 
Hope this helps


==============================================Johannes van der Pol
Post-doc, plateforme VIA-INNO, GREThA UMR-CNRS 5113
PhD consultant Labex TRAIL
University of Bordeaux
Avenue Léon Duguit, 33600 Pessac
Office F346a












Le 29 juin 2017 à 00:03, Bandeh Ali Talpur [hidden email] [netlogo-users] <[hidden email]> a écrit :


Hi All, 
Not sure if my question is relevant to this group or not. But I hope I will get some help here. 
I'm visualising millions of nodes in social network .. Bank customers based on their shared interests, joint accounts, transactions etc. 



Is there any clustering analysis tools or techniques/algorithm you may would suggest that I could apply to narrow down from large network to small network ? 


An obvious goal here is to find communities to recommend or do any predictive analysis.
Thanks ,  


Sent from Yahoo Mail for iPhone






  #yiv0734046842 #yiv0734046842 -- #yiv0734046842ygrp-mkp {border:1px solid #d8d8d8;font-family:Arial;margin:10px 0;padding:0 10px;}#yiv0734046842 #yiv0734046842ygrp-mkp hr {border:1px solid #d8d8d8;}#yiv0734046842 #yiv0734046842ygrp-mkp #yiv0734046842hd {color:#628c2a;font-size:85%;font-weight:700;line-height:122%;margin:10px 0;}#yiv0734046842 #yiv0734046842ygrp-mkp #yiv0734046842ads {margin-bottom:10px;}#yiv0734046842 #yiv0734046842ygrp-mkp .yiv0734046842ad {padding:0 0;}#yiv0734046842 #yiv0734046842ygrp-mkp .yiv0734046842ad p {margin:0;}#yiv0734046842 #yiv0734046842ygrp-mkp .yiv0734046842ad a {color:#0000ff;text-decoration:none;}#yiv0734046842 #yiv0734046842ygrp-sponsor #yiv0734046842ygrp-lc {font-family:Arial;}#yiv0734046842 #yiv0734046842ygrp-sponsor #yiv0734046842ygrp-lc #yiv0734046842hd {margin:10px 0px;font-weight:700;font-size:78%;line-height:122%;}#yiv0734046842 #yiv0734046842ygrp-sponsor #yiv0734046842ygrp-lc .yiv0734046842ad {margin-bottom:10px;padding:0 0;}#yiv0734046842 #yiv0734046842actions {font-family:Verdana;font-size:11px;padding:10px 0;}#yiv0734046842 #yiv0734046842activity {background-color:#e0ecee;float:left;font-family:Verdana;font-size:10px;padding:10px;}#yiv0734046842 #yiv0734046842activity span {font-weight:700;}#yiv0734046842 #yiv0734046842activity span:first-child {text-transform:uppercase;}#yiv0734046842 #yiv0734046842activity span a {color:#5085b6;text-decoration:none;}#yiv0734046842 #yiv0734046842activity span span {color:#ff7900;}#yiv0734046842 #yiv0734046842activity span .yiv0734046842underline {text-decoration:underline;}#yiv0734046842 .yiv0734046842attach {clear:both;display:table;font-family:Arial;font-size:12px;padding:10px 0;width:400px;}#yiv0734046842 .yiv0734046842attach div a {text-decoration:none;}#yiv0734046842 .yiv0734046842attach img {border:none;padding-right:5px;}#yiv0734046842 .yiv0734046842attach label {display:block;margin-bottom:5px;}#yiv0734046842 .yiv0734046842attach label a {text-decoration:none;}#yiv0734046842 blockquote {margin:0 0 0 4px;}#yiv0734046842 .yiv0734046842bold {font-family:Arial;font-size:13px;font-weight:700;}#yiv0734046842 .yiv0734046842bold a {text-decoration:none;}#yiv0734046842 dd.yiv0734046842last p a {font-family:Verdana;font-weight:700;}#yiv0734046842 dd.yiv0734046842last p span {margin-right:10px;font-family:Verdana;font-weight:700;}#yiv0734046842 dd.yiv0734046842last p span.yiv0734046842yshortcuts {margin-right:0;}#yiv0734046842 div.yiv0734046842attach-table div div a {text-decoration:none;}#yiv0734046842 div.yiv0734046842attach-table {width:400px;}#yiv0734046842 div.yiv0734046842file-title a, #yiv0734046842 div.yiv0734046842file-title a:active, #yiv0734046842 div.yiv0734046842file-title a:hover, #yiv0734046842 div.yiv0734046842file-title a:visited {text-decoration:none;}#yiv0734046842 div.yiv0734046842photo-title a, #yiv0734046842 div.yiv0734046842photo-title a:active, #yiv0734046842 div.yiv0734046842photo-title a:hover, #yiv0734046842 div.yiv0734046842photo-title a:visited {text-decoration:none;}#yiv0734046842 div#yiv0734046842ygrp-mlmsg #yiv0734046842ygrp-msg p a span.yiv0734046842yshortcuts {font-family:Verdana;font-size:10px;font-weight:normal;}#yiv0734046842 .yiv0734046842green {color:#628c2a;}#yiv0734046842 .yiv0734046842MsoNormal {margin:0 0 0 0;}#yiv0734046842 o {font-size:0;}#yiv0734046842 #yiv0734046842photos div {float:left;width:72px;}#yiv0734046842 #yiv0734046842photos div div {border:1px solid #666666;min-height:62px;overflow:hidden;width:62px;}#yiv0734046842 #yiv0734046842photos div label {color:#666666;font-size:10px;overflow:hidden;text-align:center;white-space:nowrap;width:64px;}#yiv0734046842 #yiv0734046842reco-category {font-size:77%;}#yiv0734046842 #yiv0734046842reco-desc {font-size:77%;}#yiv0734046842 .yiv0734046842replbq {margin:4px;}#yiv0734046842 #yiv0734046842ygrp-actbar div a:first-child {margin-right:2px;padding-right:5px;}#yiv0734046842 #yiv0734046842ygrp-mlmsg {font-size:13px;font-family:Arial, helvetica, clean, sans-serif;}#yiv0734046842 #yiv0734046842ygrp-mlmsg table {font-size:inherit;font:100%;}#yiv0734046842 #yiv0734046842ygrp-mlmsg select, #yiv0734046842 input, #yiv0734046842 textarea {font:99% Arial, Helvetica, clean, sans-serif;}#yiv0734046842 #yiv0734046842ygrp-mlmsg pre, #yiv0734046842 code {font:115% monospace;}#yiv0734046842 #yiv0734046842ygrp-mlmsg * {line-height:1.22em;}#yiv0734046842 #yiv0734046842ygrp-mlmsg #yiv0734046842logo {padding-bottom:10px;}#yiv0734046842 #yiv0734046842ygrp-msg p a {font-family:Verdana;}#yiv0734046842 #yiv0734046842ygrp-msg p#yiv0734046842attach-count span {color:#1E66AE;font-weight:700;}#yiv0734046842 #yiv0734046842ygrp-reco #yiv0734046842reco-head {color:#ff7900;font-weight:700;}#yiv0734046842 #yiv0734046842ygrp-reco {margin-bottom:20px;padding:0px;}#yiv0734046842 #yiv0734046842ygrp-sponsor #yiv0734046842ov li a {font-size:130%;text-decoration:none;}#yiv0734046842 #yiv0734046842ygrp-sponsor #yiv0734046842ov li {font-size:77%;list-style-type:square;padding:6px 0;}#yiv0734046842 #yiv0734046842ygrp-sponsor #yiv0734046842ov ul {margin:0;padding:0 0 0 8px;}#yiv0734046842 #yiv0734046842ygrp-text {font-family:Georgia;}#yiv0734046842 #yiv0734046842ygrp-text p {margin:0 0 1em 0;}#yiv0734046842 #yiv0734046842ygrp-text tt {font-size:120%;}#yiv0734046842 #yiv0734046842ygrp-vital ul li:last-child {border-right:none !important;}#yiv0734046842