|
JAIST Repository >
科学技術開発戦略センター 2003～2008 >
z2-70. JAIST PRESS 発行誌等 >
IFSR 2005 >
このアイテムの引用には次の識別子を使用してください:
http://hdl.handle.net/10119/3917
|
| タイトル: | Wrapper Feature Extraction for Time Series Classification Using Singular Value Decomposition |
| 著者: | Hui, Zhang
Tu, Bao Ho
Kawasaki, Saori |
| キーワード: | time series
feature extraction
wrapper approach
singular value decomposition |
| 発行日: | Nov-2005 |
| 出版者: | JAIST Press |
| 抄録: | Time series classification is an important aspect of time series mining. Recently, time series classification has attracted increasing interests in various domains. However, the high dimensionality property of time series makes time series classification a difficult problem. The so-called curse of dimensionality not only slows down the process of classification but also decreases the classification quality. Many dimensionality reduction techniques have been proposed to circumvent the curse of dimensionality problem for improving the time series classification performance. However, most of the proposed time series dimensionality reduction algorithms don’t utilize the information of data labels that is crucial for the classification problem. We propose a wrapper feature extraction algorithm incorporating with the classification algorithm for time series classification in this paper. The classification errors estimated by cross validation are taken as the measure of the quality of dimensionality reduction. As a set of univariate time series can be represented as a matrix, singular value decomposition is used as the feature extraction algorithm to approximate the original time series with a lower-rank matrix. By analyzing the characters of singular vectors for noisy data, we propose several efficient search algorithms. Comparison Experiments on several benchmark time series data validate the usefulness of the proposed approach. |
| 記述: | The original publication is available at JAIST Press http://www.jaist.ac.jp/library/jaist-press/index.html
IFSR 2005 : Proceedings of the First World Congress of the International Federation for Systems Research : The New Roles of Systems Sciences For a Knowledge-based Society : Nov. 14-17, 2127, Kobe, Japan
Symposium 5, Session 4 : Data/Text Mining from Large Databases Text Mining |
| 言語: | ENG |
| URI: | http://hdl.handle.net/10119/3917 |
| ISBN: | 4-903092-02-X |
| 出現コレクション: | IFSR 2005
|
このアイテムのファイル:
| ファイル |
記述 |
サイズ | 形式 |
|---|
| 20133.pdf | | 100Kb | Adobe PDF | 見る/開く |
|
当システムに保管されているアイテムはすべて著作権により保護されています。
|
