Computing reverse nearest neighbourhood on road maps

Abstract

A reverse nearest neighbour (RNN) query returns every object for which the query is its nearest neighbour. Recently, a new useful variant of the RNN query has emerged, called reverse nearest neighbourhood (RNNH) query, to discover the neighbourhood that find query point is the nearest facilities among all other fa- cilities. However, the existing research focuses on computing RNNH in Euclidean space and, to the best of our knowledge, there does not exist any technique to compute RNNH in road networks. The existing techniques designed for Euclidean space cannot be applied for road networks mainly because all objects and facilities of interest are connected on a road network. Thus, there is a need for techniques specifically designed for computing RNNH in road networks. In this paper, we introduce the reverse nearest neighbourhood concept for road networks (RNNH-RN) where a neighbourhood is a collection of at least m chained objects such that the maximum road network distance between any pair of objects is at most d. We demonstrate the flexibility and effectiveness of the proposed RNNH-RN query through extensive experiments based on real-world, road network data for Melbourne, Victoria. Our experimental results demonstrate that our solutions are feasible for networks with low, medium or high levels of object density.