Top-k Manhattan spatial skyline queries

Abstract

Data retrieval from a huge spatial database has been the subject of research fields including database systems, geographic information systems, and computational geometry for many years. In this context, we study the retrieval of relevant points with respect to a query and a scoring function: For two point sets P and Q in the plane, the skyline of P with respect to Q consists of points of P for which no other point of P is closer to all points of Q. A skyline of a point set P with respect to a query set Q can be seen as the most “relevant” or “desirable” subset of P with respect to Q. As the skyline of a set P can be as large as P itself, it is reasonable to filter the skyline further using a scoring function f that reflects the relevance of each point in the skyline well, and to report only the k best skyline points with respect to f.In this paper, we consider the top-k Manhattan spatial skyline query problem with respect to monotone scoring functions which quantifies, for each point in P, how well it fits the given query under the L1 distance. We present an algorithm that computes the top-k skyline points in time near linear in the size of P, assuming that f and k are part of the input. The presented strategy improves over the direct approach of using the state-of-the-art algorithm to compute the Manhattan spatial skyline and then filtering it by the scoring function by a log⁡(|P|) factor. Our empirical results suggest that our algorithm outperforms the direct approach by an order of magnitude.