Learning to detect representative data for large scale instance selection

NCU Institutional Repository > 管理學院 > 資訊管理學系 > 期刊論文 > Item 987654321/106821

jsp.display-item.identifier=請使用永久網址來引用或連結此文件: https://ir.lib.ncu.edu.tw/handle/987654321/106821

题名:	Learning to detect representative data for large scale instance selection
作者:	蔡志豐;Lin, Wei-Chao;Tsai, Chih-Fong;Ke, Shih-Wen;Hung, Chia-Wen;Eberle, William
贡献者:	管理學院資訊管理學系
关键词:	Algorithms;Artificial intelligence;Cost reduction;Data mining;Data processing;Data reduction;Datasets;Instance selection;Sensors;Studies
日期:	2015-01-01
上传时间:	2026-04-23 13:45:28 (UTC+8)
出版者:	Elsevier Inc.;New York: Elsevier Inc
摘要:	摘要： •The ReDD (Representative Data Detection) approach is introduced for large scale instance selection.•In ReDD, a detector learns the patterns of (un)representative data after performing instance selection.•Then, the detector is used to detect the newly added data.•We found that ReDD can reduce the computational cost and maintain the final classification accuracy. Instance selection is an important data pre-processing step in the knowledge discovery process. However, the dataset sizes of various domain problems are usually very large, and some are even non-stationary, composed of both old data and a large amount of new data samples. Current algorithms for solving this type of scalability problem have certain limitations, meaning they require a very high computational cost over very large scale datasets during instance selection. To this end, we introduce the ReDD (Representative Data Detection) approach, which is based on outlier pattern analysis and prediction. First, a machine learning model, or detector, is used to learn the patterns of (un)representative data selected by a specific instance selection method from a small amount of training data. Then, the detector can be used to detect the rest of the large amount of training data, or newly added data. We empirically evaluate ReDD over 50 domain datasets to examine the effectiveness of the learned detector, using four very large scale datasets for validation. The experimental results show that ReDD not only reduces the computational cost nearly two or three times by three baselines, but also maintains the final classification accuracy. 出版者： New York: Elsevier Inc 出版日期： 2015-08 出處： The Journal of systems and software, 2015-08, Vol.106, p.1-8 資源來源： Elsevier ScienceDirect Journals Complete 版權： 2015 Elsevier Inc. 版權： Copyright Elsevier Sequoia S.A. Aug 2015 識別號： ISSN: 0164-1212 識別號： EISSN: 1873-1228 識別號： DOI: 10.1016/j.jss.2015.04.038 識別號： CODEN: JSSODM
显示于类别:	[資訊管理學系] 期刊論文

文件中的档案:

档案	描述	大小	格式	浏览次数
index.html		0Kb	HTML	25	检视/开启

在NCUIR中所有的数据项都受到原著作权保护.

社群 sharing

数据加载中.....