Frequent subgraph mining (FSG) has always been an important issue in data mining. Several frequent subgraph mining methods have been developed for mining graph data. However, most of these are main memory algorithms in which scalability is a bigger issue. A few algorithms have opted for a relational approach that stores the graph data in relational tables. However, relational databases have their own space as well computing constraints when it comes to storing large databases. Moreover, relational databases do not preserve semantic information as they represent simple entities and in order to preserve the relationship between two entities additional tables are necessary. Object-oriented databases, on the other hand, do not have these constraints. In this paper, we present an object-oriented database approach to mining frequent sub-graphs. We use Db4o, a popular open-source object database system, to store the input graph data as well as intermediate results. Db4o can save all the information about an entity in a single class in an object form. Application domains such as protein-protein interaction data, social network data, and chemical compound structure data require mining frequent subgraphs while preserving the meaning. This paper proposes a novel idea for using object oriented database db4o to store graph data, which can support large graph data as well as preserve semantic information.
|Cite as: Srichandan, B. and Sunderraman, R. (2011). OO-FSG: An Object-Oriented Approach to Mine Frequent Subgraphs. In Proc. Australasian Data Mining Conference (AusDM 11) Ballarat, Australia. CRPIT, 121. Vamplew, P., Stranieri, A., Ong, K.-L., Christen, P. and Kennedy, P. J. Eds., ACS. 221-228
(local if available)