Top-n Reverse Nearest Neighbor Queries

• Main Authors: Ming-yi Lo, 羅銘儀
• Other Authors: Tai-lin Chin
• Format: Others
• Language: en_US
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/37184915049082048028

碩士 === 國立臺灣科技大學 === 資訊工程系 === 101 === A Reverse Nearest Neighbor query (RNN) retrieves data points at which the query point is their nearest neighbor. A generalization of RNN is the Reverse k Nearest Neighbor query (RkNN) that retrieves data points at which the query point is within their k nearest neighbors. Recently, many applications of RkNN are proposed, such as profile-based marketing, resource allocation, decision support systems, etc. For RkNN query, there is a degree of influence for the reverse relation between query point q and data point p. That is, when q is the nearest neighbor of p, it can be said that p is the most influenced by q. In this paper, a study of the Top-n Reverse Nearest Neighbor (TnRNN) query problem is proposed, where the retrieved n data points are the most influenced points of the query point. An efficient algorithm called Hexangular Search for TnRNN processing. This approach does not need to do any pre-processing, it applies the Hexangular property of RkNN, which restricts the searching range of the query, and iteratively searches the data points until n most influenced data points are found. Experiments are conducted to show that the Hexangular Search algorithm outperforms previous techniques on TnRNN query when the amount of data is large.