An incremental approach to closest pair queries in spatial networks using best-first search

Abstract

This paper addresses the problem of k Closest Pairs (kCP) query in spatial network databases. A Best-First search approach namely BFCP (Best-First Closest Pair) is proposed. Given two data sets of objects in a spatial network, BFCP first finds the 1st CP by computing the 1st NN (nearest neighbor) of each object in the set with smaller cardinality. Then BFCP retrieves the 2nd , 3rd , ⋯, kth CP in an incremental way by searching the next NN of the currently found CP's source point. Furthermore, a novel buffer replacement policy called MDU (Minimum Distance Unit) is proposed to reduce I/O cost of BFCP. Unlike LRU, which records only the last reference time, the MDU policy considers both temporal locality and spatial locality when selecting a buffer page as the victim. A comprehensive experimental study is conducted to demonstrate the advantage of BFCP and MDU. © 2011 Springer-Verlag Berlin Heidelberg.