|
| | | |
Efficiently Computing the Top N Averages in Iceberg Cubes
Chou, P.L. and Zhang, X.
Data cubes are an enabling approach for efficient on-line analytical processing (OLAP) systems. Iceberg cubes are special cubes consisting of the aggregates of multi-dimensional groups that satisfy user-specified thresholds. When there are a relatively large number of dimensions, the number of groups in an iceberg cube is huge. End users cannot fully understand the aggregate results or directly use them to make a decision. Approaches for the efficient computation of the top n groups have been proposed and have been shown to work well on iceberg cubes with simple aggregate functions such as COUNT and SUM. In this paper, we study the efficient computation of the top n groups with the complex aggregate function AVERAGE. As the average of a sub-group does not increase/decrease monotonically with its super-group, AVERAGE constraints cannot be used directly for pruning. We propose a new technique upper-bounding verage which is anti-monotonic and can be used for low-cost effective pruning. Based on a tree structure representing groups, search and pruning techniques are developed, and an algorithm is proposed to compute the top n averages efficiently. |
Cite as: Chou, P.L. and Zhang, X. (2003). Efficiently Computing the Top N Averages in Iceberg Cubes. In Proc. Twenty-Sixth Australasian Computer Science Conference (ACSC2003), Adelaide, Australia. CRPIT, 16. Oudshoorn, M. J., Ed. ACS. 101-109. |
(from crpit.com)
(local if available)
|
|