Font Size: a A A

Approximate answering of aggregate queries in relational databases

Posted on:2003-12-31Degree:Ph.DType:Dissertation
University:Georgia Institute of TechnologyCandidate:Jermaine, Christopher MatthewFull Text:PDF
GTID:1468390011489039Subject:Computer Science
Abstract/Summary:
Since hard-disk storage performance does not keep up with the ever-increasing size of these databases, approximate query answering has become a promising way to improve performance. For analytical tasks, slight errors are often acceptable if accompanied by a drastic decrease in response time. An analyst can explore the data using approximations, and verify any significant findings by issuing exact queries later on.; This dissertation explores several ways to accurately approximate the answers to statistical queries over common aggregate functions like SUM, COUNT, and AVERAGE. The dissertation discusses methods based on wavelet decompositions, kernel estimation, and sampling combined with maximum likelihood estimation. The dissertation has an emphasis on difficult cases where the data are high-dimensional, or where the data objects have many attributes (potentially hundreds).
Keywords/Search Tags:Data, Approximate, Queries
Related items