Multi-scale visualization of a collaboration dataset

> By Massoud Seifi, Jean-Loup Guillaume and Matthieu Latapy

Multi-scale visualization of a collaboration dataset

Multi-scale visualization of a collaboration dataset

Many real-world networks can be represented as large graphs. Computational manipulation of such large graphs is common, but current tools for graph visualization are limited to datasets of a few thousand nodes.

These graphs contain sets of highly connected nodes that we call “communities”. Furthermore, these communities often have their own parts which are more connected than the rest that can be viewed as “sub-communities”. We used the Louvain method to extract communities and sub-communities from a sample network obtained from Arxiv dataset. We also used GUESS which is a graph exploration tool that contains an interpreted language (Gython) combined with a graphical front-end.

Using extracted hierarchical clustering dendrogram from Louvain method, we developed a tool which visualizes different hierarchical partitions of graph. Also, it allows us to manually merge and unmerge nodes into and from a community.

The plot shows the five levels of the decomposition, the smallest graph being the one between the communitiues whose decomposition maximizes the modularity according to Louvain method.

This entry was posted in Plots and tagged ,