請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/30063
標題: | 時間序列資料庫中封閉性多序列樣式之資料探勘 Mining Closed Patterns in Multi-sequence Time-series Databases |
作者: | Tzu-Yu Lee 李梓煜 |
指導教授: | 李瑞庭 |
關鍵字: | 資料探勘,序列樣式,時間序列,封閉性樣式,演算法, data mining,sequential pattern,time-series sequence,closed pattern,algorithm, |
出版年 : | 2007 |
學位: | 碩士 |
摘要: | 序列樣式分析已經被廣泛地應用在許多領域上,且已經有許多尋找序列樣式的方法被提出。但是目前所提出的方法只考慮每筆交易只含有一條序列,並沒有考慮每筆交易含有多序列的情形,也沒有考慮到項目集合間的時間間隔,因此,這篇論文將探討尋找多序列樣式之資料探勘。在本篇論文中我們提出一個有效率的探勘演算法叫「CMP-Miner」,以找尋時間序列資料庫中封閉性多序列樣式。我們的方法可以分為三個階段。第一階段,我們將每個時間序列轉換成一個符號序列。第二階段,我們產生所有的頻繁樣式,並且在產生的過程中對這些樣式做檢查,檢查它們是否為封閉的。在第三階段,我們將重複執行第二階段的步驟直到不能找到任何的封閉性樣式為止。實驗結果顯示,我們所提出的方法具有效率與擴充性,我們的方法比Apriori演算法快達數十倍之多。 There are many algorithms proposed to find sequential patterns in a sequence database. However, the sequential pattern mining algorithms proposed are not suitable for mining frequent patterns in a time-series database since they do not consider multiple sequences in a transaction and the time intervals between the itemsets in a frequent pattern. Therefore, in this thesis, we propose an efficient algorithm, called CMP-Miner, to mine closed patterns in time-series databases where each transaction contains multiple sequences. Our proposed algorithm consists of three phases. First, we transform each time-series sequence to a symbolic sequence. Second, we generate all frequent patterns and check whether the frequent patterns are closed during the process of pattern generation. The second phase is repeated until no more closed patterns can be generated. The experimental results show that our proposed algorithm is efficient and scalable, and outperforms the modified Apriori algorithm by one order of magnitude. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/30063 |
全文授權: | 有償授權 |
顯示於系所單位: | 資訊管理學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-96-1.pdf 目前未授權公開取用 | 534.89 kB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。