Publication Type

Journal Article

Version

acceptedVersion

Publication Date

7-2008

Abstract

Given a set of data points P and a query point q in a multidimensional space, Reverse Nearest Neighbor (RNN) query finds data points in P whose nearest neighbors are q. Reverse k-Nearest Neighbor (RkNN) query (where k ≥ 1) generalizes RNN query to find data points whose kNNs include q. For RkNN query semantics, q is said to have influence to all those answer data points. The degree of q's influence on a data point p (∈ P) is denoted by κp where q is the κp-th NN of p. We introduce a new variant of RNN query, namely, Ranked Reverse Nearest Neighbor (RRNN) query, that retrieves t data points most influenced by q, i.e., the t data points having the smallest κ's with respect to q. To answer this RRNN query efficiently, we propose two novel algorithms, κ-Counting and κ-Browsing that are applicable to both monochromatic and bichromatic scenarios and are able to deliver results progressively. Through an extensive performance evaluation, we validate that the two proposed RRNN algorithms are superior to solutions derived from algorithms designed for RkNN query.

Keywords

Algorithms, Database, Nearest Neighbor, Query Processing, Reverse Nearest Neighbor

Discipline

Databases and Information Systems | Numerical Analysis and Scientific Computing

Publication

IEEE Transactions on Knowledge and Data Engineering

Volume

20

Issue

7

First Page

894

Last Page

910

ISSN

1041-4347

Identifier

10.1109/TKDE.2008.36

Publisher

IEEE

Additional URL

http://dx.doi.org/10.1109/TKDE.2008.36

Download

Share

COinS