Un partitionnement basé sur la densité de graphe pour approcher la fouille distribuée de sous-graphes fréquents.

In recent years, graph mining approaches have become very popular, especially in domains such as databases, machine learning, bioinformatics, chemoinformatics and social networks. A challenging tasks in this context is frequent subgraph discovery which has been highly motivated by the tremendously increasing size of existing graph databases. Consequently, there is an urgent need of efficient and scaling approaches for frequent subgraph mining. In this paper, we propose a MapReduce-based approach for large-scale subgraph mining. The proposed approach provides a density-based partitioning technique that take into account data characteristics and enhances the default data partitioning technique of MapReduce. Our experiments show that the proposed approach decreases significantly the runtime and scales the subgraph discovery process to large graph databases. [ABSTRACT FROM AUTHOR]