在實際的應用環境中，資料庫的內容是不斷地快速成長及改變的，因此，研究一個具有自我學習及具調適性(adaptive)的關聯法則探勘方法，以便從動態資料庫中，即時地(on-line)挖掘出使用者所感興趣的知識法則，幫助決策者機動且快速地做出正確的決策，乃是企業提昇競爭力的一個重要課題。此即關聯法則之線上探勘及漸進式探勘所要解決的問題。 此外，由於挖掘出的規則數量可能眾多，或是對使用者而言是無意義或具有重覆性的，因此必須制定度量規則有趣性的準則。在本論文中，我們提出一種以相關性縮減項目集絡 (Correlative Reduced Itemset Lattice, CRIL)為基礎之線上調適性關聯法則挖掘法，該方法可隨著門檻值的不同、資料庫的變化及項目間的相關性動態地調整CRIL，以便快速地產生使用者所感興趣的規則。我們探討本方法在實作上面對的問題及解決的方法，並且透過許多實驗結果的分析，驗証本方法的效率以及評估其優缺點和可行性。 In the real application environment, the content of databases grows quickly and incrementally; therefore, an on-line and adaptive miner, which can efficiently mine interesting association rules from the dynamic databases, is very necessary for helping the manager to make the correct decisions in order to promote the competition ability of businesses. The on-line and incremental mining of association rules are the process trying to achieve the requirements. Besides, the mined association rules may be redundant or useless to the users so an interesting measure of rules is needed to eliminate the ineffective rules. In this study, we propose an adaptive miner based on the Correlative Reduced Itemset Lattice (CRIL). The proposed approach can dynamically adjust the CRIL with the different thresholds, the variation of databases, and the correlation of the itemsets to rapidly generate the interesting rules that the users really want. Several experiments were conducted to verify the effectiveness and feasibility of the proposed association rule miner.
