Efficient top-K query algorithms using density index

Abstract

Top-k query has been widely studied recently in many applied fields. Fagin et al. [3] proposed an efficient algorithm, the Threshold Algorithm (i.e. TA), to process top-k queries. However, in many cases, TA does not terminate even if the final top-k results have been found for some time. Based on these, we propose a novel algorithm: Density Threshold Algorithm (i.e. DTA), which is designed to minimize the useless accesses of a top-k query, and introduce a novel indexing structure, Density Index, to support our algorithms. However, we proved the DTA is not instance optimal in Fagin's notion and we also propose an instance optimal algorithm named Selective-Density Threshold Algorithm (i.e. S-DTA). Finally, extensive experiments show that our algorithms have significant improvement on the efficiency, compared with the TA algorithm. © 2011 Springer-Verlag.