R. Ivancsy and I. Vajk (Hungary)
frequent pattern mining, graph-based data mining, frequent subgraph, graph-isomorphism
Frequent pattern mining in transactional databases is a highly researched area in the field of data mining. Several algorithms appeared in order to efficiently discover the fre quent itemsets in large amount of data. However not all re lationships of the real world problems can be modeled with such transactional data. More complex models are needed such as graphs. Within this model the problem of finding frequent itemsets becomes that of discovering frequent sub graphs. In recent years several graph-based frequent pat tern mining algorithms have been developed which solves different types of problems. They differ in several aspects for example in the graph representation or in the complete ness of the subgraph discovery process. Because of these differences the effectiveness of the algorithms varies signif icantly. When dealing with a real world problem the most adequate algorithm has to be chosen in order to solve the given problem in the most efficient way. For this reason it is desirable to classify these algorithms. In this paper a trade-off is illustrated, namely, which aspects of selection should be considered when one classifies graph-based fre quent pattern discovering algorithms. Brief descriptions of some approaches belonging to the different classes are pre sented as well.
Important Links:
Go Back
