Browsing into Data Masses: Graph Clustering and Visual Interpretation

Web Site: http://romain.raveaux.free.fr

A Clustering algorithm is applied on graphs and a visual interpretation is proposed.

Model-Driven Engineering (MDE) promotes the use of models as first class
citizens. As a consequence, dealing with a wide variety of modeling languages
can be a cumbersome task. Developing tools and methods for dealing with its
variety of models is a challenging task for the MDE community. In this paper,
we propose a method for identifying families of modeling languages in
megamodels. This approach makes use of graph mining methods for automatically
labeled metamodels with a partitional clustering algorithm applied on a set of
features called graph probing. Metamodels tagged with the same labels form a
“Family” what is to say sharing common substructures. Finally, a
statistical cooccurrence matrix was computed on class names and association
roles to bring a semantic meaning to each metamodel family, building somehow a
tag cloud of the family.