Exploring correlation for fast skyline computation

Boseon Yu, Wonik Choi, Ling Liu

Abstract: Skyline queries continue to attract attentions since the skyline operator was first proposed in 2001. In this paper, we propose a fast skyline computation approach, called HashSkyline, with two unique features: First, HashSkyline minimizes the pre-processing cost to O(n) by effectively utilizing the characteristics of correlated datasets. Second, HashSkyline capitalizes on a hash cell based mechanism to learn the level of correlation among the data points in a given dataset at low processing cost, allowing early detection of anti-correlated datasets at early stage to avoid uninteresting and yet high cost of computing skylines on anti-correlated datasets.