Try using graph layout algorithm for visualizing your graph so that you can understand the various  parameters regarding your graph. There are many layout types for graph are available check out those and see which suites your problem better.

Please check following article :-

http://www.ijettcs.org/Volume3Issue2/IJETTCS-2014-04-20-099.pdf

-Mayur

sure Mayur, we can tweak the visualisation in order to get simpler view about the network. But what i really want to do is reducing the number of nodes and edges in a systematic way, so i will see simpler graph ..

Dear Andry, 1) you can use various tree decompositions, 2) you can use graph homomorphisms, map graphs to simpler graphs, this may be hard for big social networks.

As you want to view your graph as reduced graph then you can view it as a clustered graph or compound graph where similar kind of nodes and edges would be treated as one cluster likewise make many clusters in your graph. Probably you can view a cluster as a single entity in your graph and not those nodes and edges belonging to that cluster. If you need those nodes and edges back after some time then you can get back those values if stored while creating clusters.

I know there are many more parameters to be considered while preparing clusters but this technique will help you to minimizer graph and view it in terms of clusters. For more info look for clustered graph layouts.

Halo Peteris, I like the idea of graph homomorphisms (i just knew it), i thought for a long time that there must be a way to transforms graph into a simpler one. i would think in very complex graph where number edges far numerous than nodes than maybe i can swap nodes to edges so its easier to compute, but as you say, graph homomorphism must be difficult for large graph. i will check on tree decomposition... thanks

Hi Mayer, thanks for the clustering idea .. i think its a good idea, for some large with many clustered network, this solution might be work ..

you can use GraphLab for mining and visualize big graph. it contains many algorithms that you can classify or cluster graph with billion nodes and edges

You can extract the core network from  big network  or filter particular patterns from the network 

Hi, very interesting problem with no straightforward solution that I know of, but first I think there are 2 kinds of summarization that you can think of:

-summarize the "global structure" of the network

-summarize the local characteristics of your network

For the first one, the community based approach is probably the best, check the map equation for instance: http://arxiv.org/pdf/0906.1405.pdf  (they have a visualization tool)

For the second one, I don't know better than previous answers. If your graph is oriented you can also use triad count to get an idea of the local properties... Or larger patterns but then it doesn't scale.

A common approach to 'simplify' the structure of very large networks is to change your representation from relationships among nodes to relationships among groups of nodes that are structurally 'similar'. You can assign all nodes that fall under the same similarity class the same label. If there are k such classes, you may then build a new network that only has k

hi Derek, thanks for the idea, .. i think i am going to look into your suggestion

Dear Sir,

To summarize the social network is a hard task. However, if you try to cluster them according to what they share you wil have a synthetic view of social network. Ii could be represented as a graph of nodes where each note is a group of surfers. Some works are done using Galois lattice and fomal concept analysis as a foundation. But I am not sure that it works well when the number of surfers is huge.