Abstract
Group nearest neighbor (GNN) queries are a relatively new type of operations in spatial database applications. Different from a traditional kNN query which specifies a single query point only, a GNN query has multiple query points. Because of the number of query points and their arbitrary distribution in the data space, a GNN query is much more complex than a kNN query. In this paper, we propose two pruning strategies for GNN queries which take into account the distribution of query points. Our methods employ an ellipse to approximate the extent of multiple query points, and then derive a distance or minimum bounding rectangle (MBR) using that ellipse to prune intermediate nodes in a depth-first search via an R$^*$-tree. These methods are also applicable to the best-first traversal paradigm. We conduct extensive performance studies. The results show that the proposed pruning strategies are more efficient than the existing methods
Originalsprog | Engelsk |
---|---|
Titel | GIS'05 Proceedings of the 13th ACM International Workshop on Geographic Information Systems : November 4-5, 2005, Bremen, Germany co-located with CIKM 2005 |
Redaktører | Cyrus Shahabi, Omar Boucelma |
Antal sider | 8 |
Forlag | Association for Computing Machinery |
Publikationsdato | 2005 |
Sider | 192-199 |
ISBN (Trykt) | 1-59593-146-5 |
DOI | |
Status | Udgivet - 2005 |
Udgivet eksternt | Ja |
Begivenhed | 13th ACM international workshop on Geographic information systems: co-located with CIKM 2005 - Bremen, Tyskland Varighed: 4 nov. 2005 → 5 nov. 2005 Konferencens nummer: 13 |
Konference
Konference | 13th ACM international workshop on Geographic information systems |
---|---|
Nummer | 13 |
Land/Område | Tyskland |
By | Bremen |
Periode | 04/11/2005 → 05/11/2005 |