- 學位論文
結合模糊理論與FP-Growth於層級架構探勘關聯規則之研究
Utilizing Fuzzy Theory and FP-Growth to Mine Fuzzy Association Rules for Hierarchy Structure
摘要
在資料探勘技術中，關聯規則的應用最為廣泛，尤其是在零售業上更為常見，其主要目的是在於找出項目與項目間的關聯性，萃取出有價值的資訊，再加以分析、預測。而在許多應用中，使用不同的概念階層來發掘關聯規則是非常有用的，在較低的概念層級所表示的關聯規則可以比較高的概念層級呈現更多資訊。對於一個決策問題而言，是需要考慮到使用者的認知與主觀判斷所產生的認知不確定性。因此，本研究基於概念層級架構，利用表格式模糊FP-Growth進行探勘，將架構中的每一個節點視為一語意變數，由資料中找出跨層級且可以用自然語言加以解釋之多階層模糊關聯規則。 最後，將實驗所得到之結果做分析，針對不同大小之資料庫與相關參數設定下，所提出之方法在運算時間與探勘出的規則數量之變化進行探討，並與Hong et al.的方法和Hu的方法進行比較分析，實驗結果顯示所提出之方法可有效縮短探勘所計算的時間，且在效能上也有不錯的表現。
並列摘要
Regarding the data mining technique, the application of association rule is used most widely especially for retail business, whose main purpose is to find out the association between items and extracts valuable information with association for further analysis and prediction. In many applications, introducing different hierarchy Structure to discover the association rules is very useful. The association rule used in a lower hierarchy concept could present more information than a higher hierarchy concept. For a decision problem, it must consider the cognition uncertainty generated by the user’s cognition and subjective judgment. Hence, this research based on the hierarchy concept structure and used the table structure fuzzy FP-Growth for mining. In the structure, each node is viewed as a linguistic variable. The multiple-level fuzzy association rules which are of cross hierarchy and can be interpreted by nature language. At last, the result derived from the experiment was analyzed and the proposed methodology was explored regarding the computational time and the quantity variation of the mined rules according to different size of databases set and the related parameter settings. The proposed method was also compared and analyzed with Hong et al.'s method and Hu's method and the experiment result indicated that the proposed method could effectively shorten the computational time of mining as well as having good performance on the efficiency.