Finding k-Dominant G-Skyline Groups on High Dimensional Data

Abstract

Skyline query retrieves a set of skyline points which are not dominated by any other point and has attracted wide attention in database community. Recently, an important variant G-Skyline is developed. It aims to return optimal groups of points. However, when data dimensionality is high, G-Skyline result has too many groups, which makes that users cannot determine which groups are satisfactory. To find less but more representative groups of points, in this paper, we propose a novel concept of $k$ -dominant G-Skyline, which first adopts $k$ -dominance to retrieve more representative points and then computes the groups not $k$ -dominated by others. In addition, we present a two-phase algorithm to efficiently compute $k$ -dominant G-Skyline groups. In the first phase, we construct a $lk$ DG structure while pruning the points never included in any $k$ -dominant G-Skyline group as much as possible. In the second phase, using $lk$ DG, we propose two efficient $k$ -dominant G-Skyline searching methods SM-P and SM-G, which generate new candidate groups from single points and ancestor groups, respectively. Our experimental results indicate that our proposed algorithms are more efficient than the baseline methods on real and synthetic data sets.