應用於分散式系統之平行循序樣本探勘

淡江大學機構典藏 > 商管學院 > 資訊管理學系暨研究所 > 學位論文 > Item 987654321/34101

jsp.display-item.identifier=請使用永久網址來引用或連結此文件: https://tkuir.lib.tku.edu.tw/dspace/handle/987654321/34101

题名:	應用於分散式系統之平行循序樣本探勘
其它题名:	Parallel sequential pattern mining on distributed system
作者:	郭家君;Guo, Jia-jyun
贡献者:	淡江大學資訊管理學系碩士班張昭憲;Chang, Jau-shien
关键词:	循序樣本探勘;負載平衡;分散式系統;資料探勘;sequential pattern mining;load balance;distributed system;data mining
日期:	2007
上传时间:	2010-01-11 04:55:17 (UTC+8)
摘要:	本研究以提昇循序樣本探勘效率為目標，發展了一套結合多部電腦的分散式探勘系統。首先，由於直接以支持度預測子任務工作量並不準確，我們提出新的工作量預測方法。而後，本研究提出了一套負載平衡演算法-ADLB，結合動態與靜態分配的特點，對於任務採取分段分配，以達成負載平衡並降低通訊負擔。此外，我們也引用並改進文獻[20]中有關演算法切換的概念，以取得更大幅度效能改善。為驗證系統效能，我們結合了16部電腦進行分散式探勘，實驗結果顯示，本系統在各種不同節點數目下均具有良好加速比，顯示其具有處理大型資料之潛力。 The research develop a distributed mining system coordinating multi computers to promote efficiency of sequential pattern mining. First, because of the inaccuracy of directly supporting sequential pattern mining. Then, the work propose a Advanced Dynamic Load Balance algorithm -ADLB. Different to previous works, ADLB divides subtask dispatches into several stages. According to different situations, the static and dynamic load balance method are applied adeptly to prevent the task partition from skew and reduce the communication overhead simultaneously. Furthermore, we also improve the performance with on the basis of citing literature 20. Choose a proper mining algorithm for each database but not apply a single algorithm for all databases with different features. In addition, we combine the sixteen computers that adoptee distributed mining. In comparison with the previous works, the experimental results shows ADLB can effectively reduce the runtime and obtain a better speed-up ratio. This result demonstrates the potentials of ADLB for mining sequential pattern in Very Large Databases.
显示于类别:	[資訊管理學系暨研究所] 學位論文

文件中的档案:

档案	大小	格式	浏览次数

在機構典藏中所有的数据项都受到原著作权保护.

TAIR相关文章

数据加载中.....