DC Field | Value | Language |
---|---|---|
dc.contributor.author | Lee, Sanghun | ko |
dc.contributor.author | Kim, Bum-Soo | ko |
dc.contributor.author | Choi, Mi-Jung | ko |
dc.contributor.author | Moon, Yang-Sae | ko |
dc.date.accessioned | 2017-07-04T02:26:36Z | - |
dc.date.available | 2017-07-04T02:26:36Z | - |
dc.date.created | 2017-06-27 | - |
dc.date.created | 2017-06-27 | - |
dc.date.created | 2017-06-27 | - |
dc.date.issued | 2016 | - |
dc.identifier.citation | INTERNATIONAL JOURNAL OF INNOVATIVE COMPUTING INFORMATION AND CONTROL, v.12, no.2, pp.419 - 431 | - |
dc.identifier.issn | 1349-4198 | - |
dc.identifier.uri | http://hdl.handle.net/10203/224582 | - |
dc.description.abstract | The k-NN search is widely used for similarity search on various time-series data such as image, text, trajectory, and biomedical data. In this paper, we address the problem of improving the performance of multi-step k-NN search on a multidimensional index. The existing multi-step k-NN search has a critical performance problem: it produces a large tolerance from a k-NN query on the index due to use of dimensionality reduction, and the large tolerance incurs a large number of candidates, which lead to severe I/O and CPU overhead. To overcome this problem, we propose a new solution, called coefficient control multi-step k-NN search (cc-kNN search in short), which uses c ・ k instead of k in a k-NN query to obtain a tight tolerance. For this, we intuitively explain why a simple operation of increasing k can produce the tight tolerance and formally prove that the cc-kNN search finds k results correctly without any false dismissal. We also define the control constant c used in the k-NN query and formally present how to construct an estimation function for determining the constant c. Experimental results show that the proposed cc-kNN search beats the existing multi-step k-NN search in the execution time as well as the number of candidates. | - |
dc.language | English | - |
dc.publisher | ICIC INTERNATIONAL | - |
dc.title | Coefficient control multi-step k-NN search in time-series databases | - |
dc.type | Article | - |
dc.identifier.scopusid | 2-s2.0-84988915084 | - |
dc.type.rims | ART | - |
dc.citation.volume | 12 | - |
dc.citation.issue | 2 | - |
dc.citation.beginningpage | 419 | - |
dc.citation.endingpage | 431 | - |
dc.citation.publicationname | INTERNATIONAL JOURNAL OF INNOVATIVE COMPUTING INFORMATION AND CONTROL | - |
dc.contributor.localauthor | Kim, Bum-Soo | - |
dc.contributor.nonIdAuthor | Lee, Sanghun | - |
dc.contributor.nonIdAuthor | Choi, Mi-Jung | - |
dc.contributor.nonIdAuthor | Moon, Yang-Sae | - |
dc.description.isOpenAccess | N | - |
dc.type.journalArticle | Article | - |
dc.subject.keywordAuthor | Multi-step k-NN search | - |
dc.subject.keywordAuthor | Similarity search | - |
dc.subject.keywordAuthor | Time-series data | - |
dc.subject.keywordAuthor | Dimensionality reduction | - |
dc.subject.keywordAuthor | Multi-dimensional index | - |
dc.subject.keywordPlus | MOVING AVERAGE TRANSFORM | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.