* News


Integrating OLAP and Ranking: The Ranking-Cube Methodology

Dong Xin
Computer Sciences department
University of Illinois at Urbana-Champaign

Wednesday, March 21, 2007

Recent years have witnessed an enormous growth of data in business, industry, and Web applications. Database search often returns a large collection of results, which poses challenges to both efficient query processing and effective digest of the query results. To address this problem, ranked search has been introduced to database systems. We study the problem of On-Line Analytical Processing (OLAP) of ranked queries, where ranked queries are conducted in the arbitrary subset of data defined by multi-dimensional selections. While pre-computation and multi-dimensional aggregation is the standard solution for OLAP, materializing dynamic ranking results is unrealistic because the ranking criteria are not known until the query time. To overcome such difficulty, we first develop a new ranking cube method that performs semi off-line materialization and semi online computation, and then extend it to high-dimensional data. Our performance studies show that Ranking-Cube is orders of magnitude faster than previous approaches. In this talk, I will also provide a brief overview of my other research work in data mining.

Administrative support: Jacky Carley (x8291)