運用語意變數探勘階層概念之模糊關聯規則

Mining Fuzzy Association Rules From Hierarchy Concept with Linguistic Variable

廖秀珊(Hsiu-Shan Liao)

指導教授 ： 龐金宗

元智大學/資訊學院/資訊管理學系/碩士(2009年)

摘要

資料庫─現今企業儲存資料的重要技術，除了用於儲存資料外，可透過資料探勘技術，挖掘隱藏於資料庫中未知的資訊。當前資料探勘技術眾多，其中以關聯規則運用最為廣泛，關聯規則主要探討商品購買的關聯性，以提供資訊給決策者參考之用。然而，現今關聯規則研究，最小支持度與最小信心水準大都採用預先指定數值，但如何給定合適之數值卻是一大難題；此外，商品在實務上種類眾多，所產生之交易商品組合也十分複雜，因此要從中挖掘出關聯規則並不容易，此時，藉由提高交易項目之階層概念，以助於找出交易資料間存在的關聯規則，企業中各階層管理者所需的訊息層面不同，故以階層概念探勘之關聯規則，可提供不同層面之訊息。

本研究以FP-Growth為基礎，結合模糊理論與階層概念，提出了模糊多層資料探勘演算法(Fuzzy Multiple-level data Mining Algorithm，FMMA)。此演算法，運用語意變數解決最小支持度與最小信心水準找尋合適數值之難題；以階層分類為概念，子階層僅需考量父階層為頻繁項目集之節點，故可減少探勘所需的時間與空間。本研究提出之演算法，使用上更為人性化且對於探勘執行時間及所需儲存空間均有改善，而探勘出的關聯規則，以自然語言方式呈現，因而較貼近人類思維且簡單容易了解。

關鍵字

資料探勘 ； 階層概念 ； 語意變數 ； 頻率樣式成長方法

並列摘要

Database is the key technology for current enterprises to store their data. It is not only used for storing information but also mining the unknown information that hidden in the database. There are various kinds of data mining technologies, but the most common one is the association rule. The association rule is mainly used to discuss the association of the purchased products in order to provide reference information for the decision-maker. According to the current researches of the association rule, minimum support and minimum confidence generally use pre-determined value as their basic requirement. However, how to identify that the value is sufficient is very tough. Besides, in fact there are tons of goods on the market, the more goods the more complicated the goods sets will be. Thus, the association rule is very difficult to discover. To this end, we raise the hierarchy concept of the transaction items to assist us to find out the association rule. In addition, the administrator of each hierarchy needs different messages. Therefore, use the association rule founded by hierarchy concept, different messages can be sent to different hierarchy.

Based on the FP-Growth, this study associates the framework of fuzzy theory and hierarchy concept to propose an approach- FMMA, Fuzzy Multiple-level data Mining Algorithm. This algorithm uses linguistic variables to solve the problem of finding out the sufficient value by using minimum support and minimum confidence. Furthermore, according to the level taxonomy concept, the sub-level only needs to consider the nods of its parent level’s frequent itemset, thus the required mining time and space can be reduced to a certain extent. The algorithm raised in this study is more humanity and more effective in enhancing the required executive time and information storage space. The mined association rule is presented by natural language which is closer to the human thought and is easier to understand.

並列關鍵字

Data Mining ； Hierarchy Concept ； Linguistic Variable ； FP-Growth

參考文獻

[20] 丁詒恩，「仲介房屋品質特性與軟體系統架構之設計」，虎尾科技大學，碩士論文，民國九十六年。

[3] Fayyad, U.M., “Data Mining and knowledge Discovery : Making Sense Out of data”, IEEE Expert, Volume 11, Issue 5, pp. 20-25, 1996.

[4] Han, J., Fu, Y., “Discovery of multiple-level association rules from large databases”, The international conference on very large databases, pp. 420–431, 1995.

[6] J. Han, J. Pei, and Y. Yin(2000), “Mining Frequent Patterns without Candidate Generation”, Proceedings of the 2000 ACM SIGMOD International Conference on Management of Data, pp. 1-12.

[8] L. C. Lee, T. P. Hong, and T. C. Wang, “Multi-level fuzzy mining with multiple minimum supports”, Expert Systems with Applications, Vol. 34 , pp. 459-468, 2008.