Approximately Searching Aggregate k-Nearest Neighbors on Remote Spatial Databases Using Representative Query Points

Abstract

AbstractAggregate k-Nearest Neighbor (k-ANN) queries are required to develop a new promising Location-Based Service (LBS) which supports a group of mobile users in spatial decision making. As a procedure for computing exact results of k-ANN queries over some Web services has to access remote spatial databases through simple and restrictive Web API interfaces, it suffers from a large amount of communication. To overcome the problem, this paper presents another procedure for computing approximate results of k-ANN queries. It relies on a Representative Query Point (RQP) to be used as a key of a k-Nearest Neighbor (k-NN) query for searching spatial data. According to the experiments using synthetic and real data (objects), Precision of sum k-NN query results using a minimal point as RQP is less than 0.9 in the most case that the number of query points is 10, and over 0.9 in the other most cases. On the other hand, Precision of max k-NN query results using a minimal point as RQP ranges 0.47 to 0.93 according to the experiments using synthetic data (objects). The experiments using real data (objects) show that Precision of max k-NN query results is less than 0.8 in case that k is 10, and over 0.8 in the other cases. From these results, it is concluded that accuracy of sum k-NN query results is allowable and accuracy of max k-NN query results is partially allowable.KeywordsData ObjectMobile UserMinimal PointNear NeighborRange QueryThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.