一個有效率的大參考序列探勘方法及 其在全球資訊網上的應用

Abstract

資料探勘最常用於商店交易資料庫和醫院中病患就診資料庫，這一類的資料庫有一個共通特性，即每個交易記錄□所包含的項目沒有先後順序。然而在應用環境中如能考慮項目間的順序，可得到許多隱藏的重要訊息。本論文根據相關規則及其延伸問題定義新的探勘模式(mining pattern)，即參考序列，並定義一個在有序交易紀錄環境的探勘問題--大參考序列探勘。本論文提出一個利用反向索引技術的快速探勘大參考序列演算法。最後討論此探勘模式在全球資訊網的應用。

Data mining is mostly used in retail database and medical record database. In this kind of database, the transactions are not sequential. However, if considering the order in practical environment, we can get some information that is unobvious but important. This paper is based on association rule mining and its extended problems to define the new mining pattern, also called reference sequence. Then we define large reference sequence mining problem, a mining issue in sequential transaction records. Moreover, we propose an efficient mining algorithm with inverted index technique, and its applications on WWW are also discussed.