教師資料查詢 | 類別: 會議論文 | 教師: 周清江 Chichang Jou (瀏覽個人網頁)

標題:Pincer-Style Maximal Sequential Pattern Mining
學年101
學期2
發表日期2013/07/22
作品名稱Pincer-Style Maximal Sequential Pattern Mining
作品名稱(其他語言)
著者Jou, Chichang; Wu, Chen-Cheng
作品所屬單位淡江大學資訊管理學系
出版者IADIS
會議名稱7th European Conference on Data Mining (ECDM 2013)
會議地點布拉格, 捷克
摘要Apriori-based sequential pattern mining algorithms use bottom-up method. They join frequent patterns with shorter length into candidate patterns with longer length, and then repeat the process until no more candidate patterns could be generated. In many applications, only frequent maximal sequential patterns (MSP), which are not a sub-sequence of any other frequent sequential pattern, are requested. In these cases, the Apriori-based algorithms will generate all sequential patterns first, and then eliminate non-maximal ones. That would perform lots of computations not directly related to the final results. For this reason, we propose the Pincer-Style Maximal Sequential Pattern Mining algorithm, PMSPM, to obtain all frequent MSPs by eliminating most of the intermediate steps in the Apriori-based algorithms. Like a pincer’s movement, PMSPM alternates bottom-up and top-down directions to find many MSPs in the early top-down stages. Thus, minings in the bottom-up direction could safely skip many repetitive procedures. PMSPM could save lots of support counting efforts to reduce computing time. We implement PMSPM and compare its performance with that of an Apriori-like algorithm. We also test effects of database parameters on its performance.
關鍵字Sequential Pattern Mining
語言英文
收錄於
會議性質國際
校內研討會地點
研討會時間20130722~20130724
通訊作者Jou, Chichang
國別
公開徵稿Y
出版型式電子版
出處pp.79-83
相關連結
Google+ 推薦功能,讓全世界都能看到您的推薦!