K-nn Queries in Graph Databases Using M-Trees

摘要

Metric trees (m-trees) are used to organize and execute fast queries on large databases. In classical schemes based on m-trees, routing information kept in an m-tree node includes a representative or a prototype to describe the sub-cluster. Several research has been done to apply m-trees to databases of attributed graphs. In these works routing elements are selected graphs of the sub-clusters. In the current paper, we propose to use Graph Metric Trees to improve k-nn queries. We present two types of Graph Metric Trees. The first uses a representative (Set Median Graph) as routing information; the second uses a graph prototype. Experimental validation shows that it is possible to improve k-nn queries using m-trees when noise between graphs of the same class is of reasonable level.