可篩除項目集探勘是一個從製造業生產規劃所衍生出來的問題。以往的可篩除項目集探勘研究主要關注靜態商品資料庫，然而在現實世界中，商品資料庫可能會因為商品的更動而不斷地更新，因此為了維護可篩除項目集的當前狀態，傳統的可篩除項目集探勘方法需要重新掃描更新後的資料庫，以再次執行整個迭代的探勘過程，從而花費了大量的計算時間。過去，針對商品插入的可篩除項目集探勘已經被設計。在本文中，我們進一步考慮了另外兩個維護問題，即商品刪除和商品修改。我們提出了兩個方法來解決它們，第一個方法是針對商品刪除與修改之基於快速更新演算法的可篩除項目集探勘演算法，此方法將所有項目集分成幾個案例，並利用已有的資訊快速地維護最終可篩除項目集，每個案例都將採用相應的程序來減少重新掃描原始資料庫的次數，從而節省了探勘時間。第二個方法基於準可篩除項目集，對商品刪除和修改維護正確的可篩除項目集結果，這個方法旨在通過進一步減少重新掃描原始資料庫的次數來提升第一個方法的效率。當刪除/修改的商品數量與原始商品資料庫的商品總數的比率小於一定程度時，第二個方法的效能會比第一個方法的效能好很多。最後，在各種參數設置下進行實驗，以將所提出的方法與傳統的META方法進行比較。實驗結果顯示，我們提出的兩個方法在維護效率上均優於META算法，而第二個方法在三者中表現最佳。

Mining erasable itemsets is a problem derived from the production planning of the manufacturing industry. Previous erasable-itemset mining methods most focused on static product databases. In the real world, however, a product database may be continuously updated due to the change of products. Therefore, to maintain the current state of erasable itemsets, the traditional erasable-itemset mining methods need to rescan an updated database for performing the entire iterative mining process again, thus spending much calculation time. In the past, the erasable-itemsets mining with product insertion has been designed. In this thesis, we further consider the other two maintenance problems, i.e. product deletion and modification. We propose two methods to solve them. The first one is a FUP-based erasable-itemset mining algorithm for product deletion and modification. It partitions all the itemsets into several cases and uses the existing information to maintain the final erasable itemsets quickly. Each case will take the corresponding process to reduce the number of times required for rescanning the original database. The second one is based on quasi erasable itemsets to maintain the correct results for product deletion and modification. It aims to improve the efficiency of the first approach by further reducing the number of times required for rescanning the database. When the ratio of the number of deleted/modified products over the total number of products in the original database is less than a certain degree, the performance of the second method will be much better than that of the first one. Finally, experiments are made under various parameter settings to compare the proposed methods with the traditional META approach. The experimental results show that both our proposed methods outperform the META algorithm in maintenance efficiency, and the second method is the best among the three ones.

論文審定書 i
致謝 ii
摘要 iii
Abstract iv
Contents vi
List of Tables viii
List of Figures x
Chapter 1 Introduction 1
1.1 Background and Motivation 1
1.2 Thesis Organization 3
Chapter 2 Review of Related Work 4
2.1 Mining Erasable Itemsets 4
2.2 Incremental Database Environment 5
2.3 FUP Algorithm 6
2.4 Pre-Large Itemsets 9
Chapter 3 Problem Definition for Erasable-Itemset Mining 13
Chapter 4 FUP-Based Erasable-Itemset Mining Algorithm for Product Deletion and Modification 15
4.1 The Idea 15
4.2 The Notation 20
4.3 The Proposed Algorithm for Product Deletion 22
4.3.1 The Algorithm 22
4.3.2 Example 1 25
4.4 The Proposed Algorithm for Product Modification 31
4.4.1 The Algorithm 31
4.4.2 Example 2 34
Chapter 5 Quasi-Erasable-Itemset Mining Algorithm for Product Deletion and Modification 40
5.1 The Idea 40
5.2 The Notation 47
5.3 The Proposed Algorithm for Product Deletion 49
5.3.1 The Algorithm 50
5.3.2 Example 3 54
5.4 The Proposed Algorithm for Product Modification 62
5.4.1 The Algorithm 62
5.4.2 Example 4 66
Chapter 6 Experiments and Analysis 74
6.1 Experimental Environment 74
6.2 Experimental Settings 74
6.3 Experimental Results 76
6.3.1 Results of Continuous Deletions/Modifications 76
6.3.2 Results of Different Maximum Erasable Ratios 83
6.3.3 Results of Different Database Sizes 89
6.3.4 Results of Different Product Lengths 92
6.3.5 Results of Different Quasi-Erasable Parameters 95
Chapter 7 Conclusions and Future Works 98
7.1 Conclusions 98
7.2 Future Works 99
References 100

[1] R. Agrawal, T. Imielinski and A. Swami, “Mining association rules between sets of items in large databases,” The 1993 ACM SIGMOD International Conference on Management of Data, pp. 207-216, 1993.
[2] R. Agrawal and R. Srikant, “Fast algorithm for mining association rules,” The 20th International Conference on Very Large Data Bases, pp. 487-499, 1994.
[3] D. W. Cheung, J. Han, V. T. Ng, and C. Y. Wong, “Maintenance of discovered association rules in large databases: An incremental updating approach,” The Twelfth International Conference on Data Engineering, pp. 106-114, 1996.
[4] Z. H. Deng, G. Fang, Z. Wang and X. Xu. “Mining erasable itemsets,” The 2009 International Conference on Machine Learning and Cybernetics, Vol. 1, pp. 67–73, 2009.
[5] Z. H. Deng and X. R. Xu, “An efficient algorithm for mining erasable itemsets” The Sixth International Conference on Advanced Data Mining and Applications, pp. 214–225, 2010.
[6] Z. H. Deng and X. R. Xu, “Fast mining erasable itemsets using NC_sets,” Expert Systems with Applications, Vol. 39, Issue 4, pp. 4453–4463, 2012.
[7] J. Han, J. Pei, Y. Yin and R. Mao, “Mining frequent patterns without candidate generation: a frequent-pattern tree approach,” Data Mining and Knowledge Discovery, Vol. 8, Issue 1, pp. 53-87, 2004.
[8] T. P. Hong, C. Y. Wang and Y. H. Tao, "A new incremental data mining algorithm using pre-large itemsets," Intelligent Data Analysis, Vol. 5, No. 2, pp. 111-129, 2001.
[9] T. P. Hong, K. Y. Lin, C. W. Lin and B. Vo, “An incremental mining algorithm for erasable itemsets,” IEEE International Conference on Innovations in Intelligent Systems and Applications, pp. 286-289, 2017.
[10] T. P. Hong, L. H. Chen, S. L. Wang, C. W. Lin and B. Vo, “Quasi-Erasable Itemset Mining,” IEEE International Conference on Big Data, pp.1816-1820, 2017.
[11] T. Le, B. Vo and F. Coenen, “An efficient algorithm for mining erasable itemsets using the difference of NC-Sets” The 2013 International Conference on Systems, Man, and Cybernetics, pp. 2270–2274, 2013.
[12] T. Le and B. Vo, “MEI: an efficient algorithm for mining erasable itemsets” Engineering Applications of Artificial Intelligence, Vol. 27, pp.155–166, 2014.
[13] T. Le, B. Vo and G. Nguyen. “A survey of erasable itemset mining algorithms,” Data Mining and Knowledge Discovery, Vol. 4, Issue 5, pp. 356-379, 2014.
[14] G. Lee, U. Yun, H. Ryang and D. Kim, “Erasable itemset mining over incremental databases with weight conditions,” Engineering Applications of Artificial Intelligence, Vol 52, pp. 213-234, 2016.
[15] C. W. Lin, T. P. Hong and W. H. Lu, “Using the structure of pre-large trees to incrementally mine frequent itemsets,” New Generation Computing, Vol. 28, Issue 1, pp. 5-20, 2010.
[16] C. W. Lin and T. P. Hong, “Maintenance of pre-large trees for data mining with modified records,” Information Sciences, Vol. 278, pp. 88-103, 2014.
[17] C. W. Lin, G. C. Lan and T. P. Hong, “An incremental mining algorithm for high utility itemsets,” Expert Systems with Applications, Vol. 39, Issue 8, pp. 7173-7180, 2012.
[18] B. Nath, D. K. Bhattacharyya and A. Ghosh, “Incremental association rule mining: a survey,” Data Mining and Knowledge Discovery, Vol. 3, Issue 3, pp. 157-169, 2013.
[19] G. Nguyen, T. Le, B. Vo and B. Le, “A new approach for mining top-rank-k erasable itemsets,” The Sixth Conference on Intelligent Information and Database Systems, pp. 73–82, 2014.
[20] B. Vo, T. Le, T. P. Hong and B. Le, “Maintenance of a frequent-itemset lattice based on pre-large concept,” The Fifth International Conference on Knowledge and Systems Engineering, pp. 295-305, 2013.
[21] B. Vo, T. Le, T. P. Hong and B. Le, “An effective approach for maintenance of pre-large-based frequent-itemset lattice in incremental mining,” Applied Intelligence, Vol. 41, Issue 3, pp. 759-775, 2014.
[22] IBM Quest Data Mining Project. Quest Synthetic Data Generation Code. Available from http://www.almaden.ibm.com/cs/quest/syndata.html, 1996.
[23] Frequent itemset mining dataset repository, http://fimi.ua.ac.be/data/, 2012.