臺灣博碩士論文加值系統

資料探勘要應用在各個領域，必定會有該領域的決策者僅想針對某些商品作探勘。過去演算法均針對資料庫掃描後，列出所有的關聯式規則回應給使用者。為了加速使用者找尋想要的關聯式規則，我們放棄了傳統列出所有關聯式規則的方式，僅針對使用者感興趣的項目作搜尋，這樣能確保探勘結果都與使用者感興趣的項目相關，並且大量減少不相干的準則。

我們另外透過Data Cube的方式來處理item發生頻率與規則產生。優點是能夠以批次方式處理資料，而不需要將所有資料一次載入記憶體處理，能單筆資料處理後，直接將發生頻率累計，透過這樣的方式，我們僅需要掃描一次資料庫即可得到所有發生頻率，有效降低IO與記憶體使用量。並且配上Filter作前處理，以加快處理速度。最後累計頻率可直接產生規則回應給使用者。我們也透過不同的實驗參數來驗證此方法效能。

If data mining is going to apply in every domain, there must be some policy maker want to do some mining in some certain products. In former algorithm, computer would scan the entire database, and list all the association rules to user. In order to increase the pace for users to search for the association rules, we give up the traditional way to search all the association rules. We only focus on certain items which are interesting to users. Therefore, we can make sure the results would fit the need of users, and decrease a large number of un-related rules.

We use the way of Data Cube to deal with the items’ frequencies and rule generation. The advantage is to conduct the data with batch way rather than load all the data into the memory and then search them. We can conduct with every item and add up the frequencies. Through this way, what we do is only scan the database once and we can get every frequency. Efficiently reduce the IO and the use of memory. Collocating with Filter to pre-deal the data would increase the pace. In the end, the add-up frequencies can generate association rules to the user. We can verify the efficiency of this way through many different experiment parameters.

目錄
中文摘要 i
Abstract ii
第一章節 簡介 1
第二章節 相關研究與預備定理 5
Define 1：符號說明 5
Define 2：支持度Support / 信賴度 Confident 5
Define 3：Objective-Oriented utility-based Association mining (OOA) 7
Define 4：Data Cube 9
第三章節 我們的方法 11
3-1 流程簡介 11
3-2 詳細步驟說明 11
3-3 序列資料分析 14
第四章 實驗與分析 16
4-1 演算法: 16
4-2 實驗環境 17
4-3 參數分析 17
4-4 比較結果 20
第五節 結論 24
第六節 參考文獻： 25

[1]Rakesh Agrawal and Ramakrishnan Srikant, "Fast Algorithms for Mining Association Rules in Large Databases," Proceedings of the 20th International Conference on Very Large Data Bases, p.487-499, September 12-15, 1994
[2]Rakesh Agrawal and Ramakrishnan Srikant, "Mining Sequential Patterns," Proceedings of the Eleventh International Conference on Data Engineering, p.3-14, March 06-10, 1995
[3]Rakesh Agrawal , Tomasz Imielinski, and Arun Swami, “Mining association rules between sets of items in large database,” Proceedings of the 1993 ACM SIGMOD International Conference on Management of Data, Washington, D.C., May 26-28
[4] Jiawei Han and Micheline Kamber, “Data Mining Concepts and Techniques,” Morgan Kauffman, 2ed, 2006.
[5] Yi-Dong Shen, Zhong Zhang and Qiang Yang, “Objective-oriented utility-based association mining,” Proceedings of the 2002 IEEE International Conference on Data Mining, Maebashi City, Japan, December 2002, 426-433.
[6] Hong Yao, Howard J. Hamilton and Liqiang Geng, “A unified framework for utility based measures for mining itemsets,” Proceedings of ACM SIGKDD 2nd Workshop on Utility-Based Data Mining, Philadelphia, Pennsylvania, pp. 28-37, Aug. 2006.
[7] Fan Wu, Shih-Wen Chiang and Jiunn-Rong Lin, “A new approach to mine frequent patterns using item-transformation methods,” Information Systems, v.32 n.7, p.1056-1072, November, 2007
[8] Liqiang Geng and Howard J. Hamilton, “Interestingness measures for data mining: A survey,” ACM Computing Surveys (CSUR) Volume 38 , Issue 3 (2006) , Article No. 9,
[9] Jiawei Han, Jiam Pei, Yiwen Yin and Runying Mao, “Mining frequent patterns without candidate generation,” Proceedings of ACMSIGMOD Conference on Management of Data, pp. 1–12.
[10] Jiawei Han, Laks V. S. Lakshmanan and Jian Pei, “Scalable Frequent-Pattern Mining Methods: An Overview,” Proceedings of the seventh ACM SIGKDD international conference on Knowledge discovery and data mining, August 26-29, 2001, San Francisco, California
[11] Raymond Chan , Qiang Yang and Yi-Dong Shen, “Mining High Utility Itemsets,” Proceedings of the Third IEEE International Conference on Data Mining, p.19, November 19-22, 2003
[12] Ramakrishnan Srikant and Rakesh Agrawal, “Mining Sequential Patterns: Generalizations and Performance Improvements,” Preceedings of the 5th International Conference on Extending Database Technology (EDBT), Avignon, France, IBM Research Division, pp. 3-17, 1996.
[13] http://www.almaden.ibm.com/cs/disciplines/iis/