[[alternative]]An Integrated Framework with Interactively and Adaptively Mining Association Rules

[[abstract]] 過去，大部分的研究通常簡化知識發掘的過程以方便進行研究，或把挖掘過程遇到的問題分開單一討論與解決；但現實上整個程序必須整合人工處理程序及機器處理程序，而且整個知識發掘過程是一個互動且繁覆的過程，在本論文我們稱之為擴張的知識發掘模式(Intension Mining Model for KDP)，在這整個過程中所遇到的問題，我們必須一併來討論及解決才有意義。我們以資料挖掘中最受到注目的關聯法則挖掘來探討，在擴張的知識發掘模式中我們進行關聯法則挖掘時會遇到各種問題，例如：漸進式挖掘、線上挖掘、使用者有趣性問題…等。而這些問題目前有非常多的研究提出解決方式，但沒有人提出能一併解決的方法，因此我們提出一新演算法QARM (Query-based Association Rule Miner)及結合我們先前所提出的MUM(Multi-layer Update Miner)演算法，設計出一整合式構架，稱之為QAMUM(Query-based Adaptive MUM)。它能針對企業或研究需要，進行一個或多個問題的解決，這使得關聯法則挖掘更具資料合用性與實用性議題。 [[abstract]] In the past, the traditional model for knowledge discovery process (KDP) was usually simplified to facilitate the proceeding of the research, or only a single sub-problem in the KDP was solved at a time. Recently, the researchers and practitioners have realized the limitations of the traditional model and felt the need for standardization of the KDP. It is significantly meaningful that all the problems in the KDP should be considered and solved together. This is so called the Intension Mining Model for KDP.In this study, we discuss the extended model for association rule mining that is one of the popular research areas in data mining. A number of techniques and algorithms have been proposed to mine the association rules from different aspects, respectively, such as the on-line mining, the incremental mining, the interestingness problems, and so on. But few of researches have attempted to solve these problems in an integrated way. Therefore, we design a new algorithm, namely the QARM (Query-based Association Rule Miner), based on our previously proposed MUM (Multi-layer Update Miner), to construct an integrated framework, namely the QAMUM (Query-based Adaptive MUM), for mining association rules efficiently. Many experiments were conducted to verify the practicability and feasibility of the proposed approach.