re: network graph clustering based on attribute

[algotr8der]

Hello,

I was wondering if there is a clustering algorithm, layout plugin etc that is intended for attribute based clustering.

I am trying to display a network graph of a electrical network. I don't care about extracting hidden relationships. My graph is structured such that every node is a electrical bus and each edge is either a physical transmission line or a virtual connection between buses. I have node and edge attributes that provides details on which zone/area the node/edge belongs to. Zones/areas are associated with geographic regions but there are no lat/lon's associated with the data. Its impossible to track down the lat/lon of 70k buses in a electrical system.

I want to see a network with nodes and edges such that there are no overlappings and lines and node sizes are determined by a voltage attribute, while colour is determined by zone/area. Since the network is large I want to be able to zoom/pan in/out and navigate through the web of connections.

This seems like a very simple task but I can't seem to get this to work in gephi. Instead the current layout algorithms seem overkill and try to do things unnecessary for my purpose. Perhaps gephi is not designed for this task?

An example of something I want can be something like the following (without directed connections - I want a undirected graph):

http://arborjs.org/halfviz/#/mystery-of-the-secret-room

The problem with the above example is that it fits only in the allocated workspace. I want the network to be able to exist outside the boundaries of the browser window space/canvas and be able to navigate around and perhaps search for nodes/edges and be able to re-center based on search criteria. Having 70k nodes fit in such a small space would render the graph meaningless since you can't see anything.

In any case I would greatly appreciate your feedback. Thank you.

[seinecle]

I am working on such an algo - stay tuned on twitter at @seinecle.