Title: cwkNN-based approach to adaptive cost model

Authors: Zhining Liao, Hui Wang, David H. Glass

Addresses: School of Computing and Mathematics, University of Ulster, Northern Ireland BT37 0QB, UK. ' School of Computing and Mathematics, University of Ulster, Northern Ireland BT37 0QB, UK. ' School of Computing and Mathematics, University of Ulster, Northern Ireland BT37 0QB, UK

Abstract: In the environment of data integration over the internet, three major factors affect the cost of a query: network congestion situation, server contention states (workload) and data/query complexity. In Liao et al. and Liu and Liao, network congestion situation and server contention states on the cost of queries cost models are investigated and a set of formulae for server contention state are constructed using multiple regression. But in a dynamic environment, the server contention states may change over time. In order to reflect the change of sever contention states in a cost model, it would be desirable if the cost model can be updated dynamically without the cost model being reconstructed from scratch. For this reason, a tree-structured index for a novel kNN method is proposed, named cwkNN (counting weighted kNN), which can be used to classify the cost of sample queries that are not in the original server contention state. Experimental results show that our cwkNN-by-tree method works quite well in maintaining accurate cost models in a dynamic environment.

Keywords: adaptive cost models; data mining; kNN algorithms; query optimisation; data integration.

DOI: 10.1504/IJSOI.2006.011012

International Journal of Services Operations and Informatics, 2006 Vol.1 No.3, pp.203 - 220

Published online: 03 Oct 2006 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article