Integrating OLAP and Ranking: The Ranking-Cube Methodology

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. Its complete life cycle, including cube construction, incremental maintenance, and query processing, will also be discussed. Our performance studies show that Ranking-Cube is orders of magnitude faster than previous approaches.

Speaker Details

Dong Xin is a graduating Ph.D. student in the Department of Computer Science at the University of Illinois at Urbana-Champaign. He has been working in the data mining research group, directed by Professor Jiawei Han. His research interests include data mining, data warehousing, and database systems. During his Ph.D. study, he has authored or co-authored 10 research papers (published or accepted) in SIGMOD, VLDB, and KDD. He received best student paper runner-up awards in KDD 2005 and KDD 2006. Dong interned with the DMX group in summer 2005 and 2006.

Date:
Speakers:
Dong Xin
Affiliation:
University of Illinois
    • Portrait of Dong Xin

      Dong Xin

    • Portrait of Jeff Running

      Jeff Running