Social Network Analysis using Graph


(Jennifer Young) #1


Something I'd be super interested in is doing social network analysis using Graph. Currently, it seams more geared for doing link analysis (if that). I'd want to be able to do the kinds of metrics like centrality scores, etc, with the graph API. The kind of data I'm working with should do well with Graph, and it'd be wonderful to be able to do this kind of analysis right in Elastic.

Anyone else have interest in that?

Anyone at Elastic think it could be on the roadmap?


(Mark Harwood) #2

I like the framework Gephi has for a workspace with plugins for various centrality measures etc. It would be good to support that type of model and I'm tidying up the internal classes used in the workspace in anticipation of that future growth. I've yet to see a standard emerge in Javascript land for representing nodes and edges etc in a consistent fashion that would allow for plug-and-play reuse of existing algorithms. In Java land I think projects like Jung were an attempt to do that.
However, the focus right now is on simplicity of UI and we don't want the default configuration to be as bewildering as Gephi to the casual user. The richer style of interface would always have to be something achieved by enabling more advanced plugins.
Note it would also be technically possible to add a connector to Gephi that talked to the backend elasticsearch Graph API so you could make use of all the features in Gephi today. As Shay mentioned at elasticon - I had a setup like this running a year ago using the JEST client inside a Gephi plugin.

(Chirag) #3

I agree @jenoldyoung and I have already started to put my usecase for the same in place. with respect to roadmap
I think ELK addresses the requirement. for specifics like centrality i think we will need to build right plugins in the stack.


(system) #4