關聯規則的資料探勘在現今的決策輔助系統上面佔有重要的地位。在過去，決 策輔助系統的後端通常是超大型交易資料庫，已有許多相關研究致力於改善關 聯規則挖掘的效率。近年來，許多決策輔助系統開始將後端平台由傳統的超大 型交易資料庫轉移到多維度的資料倉儲系統上面。資料倉儲系統通常集中由中 央管理維護，所以許多使用者在進行資料挖掘的動作時，都是由統一的資料倉 儲系統來提供歷史性的資訊。在多維度以及多層次的資料庫綱要下，如何能夠 迅速回應來自各個不同需求的人員的關聯規則挖掘請求將是重要的議題。在這 些關聯規則探勘當中我們會發現，同樣的計算容易被提出許多次。這篇論文提 出一種多維度以及多層次的資料挖掘系統，將多個關聯規則挖掘請求視為一個 工作量，對此工作量內的各個請求進行拆解、分析並且重新安排其完成的先後 次序，利用各挖掘請求間的相似部份加以管理、重複利用，以改善整體工作量 的效益，使挖掘結果能夠更快呈獻給使用者。 Association rule mining plays an important role in decision support systems, it finds interesting rules from a huge amount of historical data. In the past when decision support systems used transactional databases as backends, researches focus on the performance improvement for mining association rules. Nowadays, decision support systems often comes with several frontends and a data warehouse as the backend; the frontends send preprocessed user queries and then fetch the requested data from the warehouse while the central data warehouse has to respond a series request from different users, answering historical data in multiple dimensions and levels. Efficiently answer mining queries on different dimensions and different levels of abstraction is an important issue for decision support systems. Based on some observations, we see that an analysis process includes a series of related queries and many mining queries share common computation results. We proposed an association rule mining system framework which processes queries as a workload, managing and optimizing materialized tables, reusing the result among queries to complete the entire workload efficiently.
