Telescope: Zooming to interesting skylines

Abstract

As data of an unprecedented scale are becoming accessible, skyline queries have been actively studied lately, to retrieve "interesting" data objects that are not dominated by any other objects, i.e., skyline objects. When the dataset is high-dimensional, however, such skyline objects are often too numerous to identify truly interesting objects. This paper studies the "curse of dimensionality" problem in skyline queries. That is, our work complements existing research efforts to address this "curse of dimensionality", by ranking skyline objects based on user-specific qualitative preference. In particular, Algorithm Telescope abstracts skyline ranking as a dynamic search over skyline subspaces guided by user-specific preference with correctness and optimality guarantees. Our extensive evaluation results validate the effectiveness and efficiency of Algorithm Telescope on both real-life and synthetic data. © Springer-Verlag Berlin Heidelberg 2007.