Bilkent University
Department of Computer Engineering

---

S E M I N A R

 

Network Partitioning for Aggregate Queries

 

Engin Demir

Ph.D Student
Computer Engineering
Bilkent University

Recently, spatial networks gain the attention of many researchers since the connectivity relation among nodes that is not explored in depth in spatial databases is much more important when compared with the geometric proximity of nodes. Graph structure is commonly used for representing spatial networks, and several algorithms are performed on graph structures to provide efficient query processing. However aggregate queries, i.e. searching for multiple adjacent nodes, dominate the computation in case of path evaluation queries and graph has deficiencies to model aggregate queries. This motivation let us introduce the concept of hypergraph models to determine the I/O cost of queries by the help of network usage statistics. Experimental results show that our hypergraph models correctly decode the I/O cost of queries based on the assumption that selected period of usage statistics resemble the upcoming usage behavior.

 

DATE: May 16, 2005, Monday @ 16:40
PLACE: EA 409