Efficient computation of Top-k G-Skyline groups on large-scale database

Abstract

The G-Skyline query aims to return the best groups that are not g-dominated by any other group of equal size; this approach plays an important role in many fields, such as multiobjective decision-making. The top-k G-Skyline query has important implications since the output size of the G-Skyline query is often too large to make decisions. Due to the massive number of candidate groups and the high index construction consumption, the existing top-k G-Skyline algorithms cannot handle large-scale databases well. In this paper, an efficient algorithm is proposed called the TGPE (Top-kG-Skyline algorithm based on Presorting and Enumeration) to rapidly compute the top-k G-Skyline groups for large-scale databases. TGPE proposes an efficient scanning method based on a presorted table to obtain tuples making up G-Skyline groups and their dominant relationships. TGPE does not need to repeatedly reconstruct the index for different skyline criteria. Three computational lemmas are presented based on sorting information and monotonicity, and they are utilized to reduce the number of candidate groups to obtain results rapidly. Finally, extensive experiments on synthetic and real datasets verify that TGPE has significant performance advantages over those of the existing algorithms.