資料挖掘為近年來客戶資源管理、行銷、醫學及其他許多領域中，將使用者有興趣的資料擷取出來的一門學科，而資料擷取的方式亦分為分群法、分類法、關聯法則擷取法等相關的方式。 在資料挖掘的學術領域中，最早被提出來的演算法即為Apriori演算法，它利用反覆讀取資料庫中關聯性資料的方式，將資料擷取出來，取得屬性之間的關聯法則，但對於現實生活中動態新增的資料庫而言，Apriori面臨了三個主要的問題，其一為大型資料庫挖掘曠日耗時，且無法即時更改支持度；二為無法有效解決資料庫中動態新增資料挖掘的問題，第三為相對於整體資料庫，較新時間點的敏感性資料無法做一個有效的擷取。 本篇論文提出兩種新的演算方式，其一為MUM (Multidimensional Update Mining)演算法，可在動態資料庫中即時挖掘出項目之間的關聯法則，以改善Apriori演算法在資料挖掘時效率不佳及無法動態挖掘新增資料庫的問題。另一種是SUM (Sensitivity Update Mining) 演算法，該演算法建立在MUM之上，用以即時挖掘敏感性資料的關聯法則，改進以往演算法必須批次處理大筆資料的問題。此兩種演算法對於目前動態新增的資料庫挖掘，提供了一個有效率的動態即時敏感性資料挖掘方式，避免敏感性資料因為無法在短期間內達到使用者設定的最小支持度而被捨棄的問題。 Data mining is the exploration and analysis of large quantities of data in order to discover meaningful patterns and rules. It is an important discipline that has widely applied in fields ranging from customer relationship management to marketing, and medicine. The discovery of association rules is an important task in data mining. The Apriori algorithm is the most popularly and widely used technique for mining association rules. However, the Apriori algorithm must scan the databases many times to discover the large itemsets so that it has three main disadvantages: (1) it is time-consuming; (2) it is not suitable for mining of incrementally growing databases due to the need of rescanning the original databases; and (3) as databases grow, the sensitive information in the new transactions can not mined effectively. In this thesis, we propose the Multidimensional Update Mining (MUM) algorithm, which does not need to rescan the original database, to mine the association rules for the incrementally growing databases. Based on the MUM algorithm, a Sensitivity Update Mining (SUM) algorithm is designed to mine sensitive information from the newly inserted transactions. Many experiments and related analysis are conducted to validate our proposed approaches.
