| Autor: |
Collet, Pierre, Tomassini, Marco, Ebner, Marc, Gustafson, Steven, Ekárt, Anikó, Peltzer, Jason B., Teredesai, Ankur M., Reinard, Garrett |
| Zdroj: |
Genetic Programming (9783540331438); 2006, p49-60, 12p |
| Abstrakt: |
Speed, cost, and accuracy are crucial performance parameters while evaluating the quality of information and query retrieval within any Database Management System. For some queries it may be possible to derive a similar result set using an approximate query answering algorithm or tool when the perfect/exact results are not required. Query approximation becomes useful when the following conditions are true: (a) a high percentage of the relevant data is retrieved correctly, (b) irrelevant or extra data is minimized, and (c) an approximate answer (if available) results in significant (notable) savings in terms of the overall query cost and retrieval time. In this paper we discuss a novel approach for approximate query answering using Genetic Programming (GP) paradigms. We have developed an evolutionary computing based query space exploration framework which, given an input query and the database schema, uses tree-based GP to generate and evaluate approximate query candidates, automatically. We highlight and discuss various avenues of exploration and evaluate the success of our experiments based on the speed, cost, and accuracy of the results retrieved by the re-formulated (GP generated) queries and present the results on a variety of query types for TPC-benchmark and PKDD-benchmark datasets. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Supplemental Index |
| Externí odkaz: |
|
