現今社會中，大部份的文件資料都會被化為數位型式置放於網路上，因此可以把網路視為一個資料庫，而且是資料量與提供者最多的資料庫，如何去挖掘這個龐大的資料庫一直是熱門的研究主題。在廣大的資料庫中去根據使用者的行為興趣來找到最合適的文章並推薦給使用者，以有許多的各種推薦的演算法被提出來解決並各有其優缺點，本論文主要目的為解決因使用者輸入資訊不足而無法做出正確的推薦的問題。
	本研究以推薦系統的方式來幫助使用者更容易的檢索出所需的文章，以文章中的關鍵字作為代表該文章的特徵向量，透過模糊關聯係數找出文章間的關聯性，並以使用者點閱過的文章順序做為輸入資料，找出與這些使用者感興趣的文章具有高度相關的其他文章推薦給使用者。我們採取的方法從一開始先使用樣本模糊相關係數找出與使用者點選的第一篇文章相關的文章後進行推薦，接著隨著使用者點選更多文章來獲得更多使用者的輸入資料後，再利用多重模糊相關係數運算找出與這些文章關聯性高的其他可能文章並推薦，透過這樣循環式的推薦，來提高推薦的正確性。
	根據本論文的實驗證實，使用此種將模糊相關係數應用於文章推薦問題的新方法，比起用傳統計算文章相關性的方法來處理文章推薦問題更加的適合。

Due to the growth of Internet , more and more content can be accessed on Internet. Finding articles which user may be interesting and recommend to user by user's behavior becomes more and more important. Many kinds of algorithm for solving this problem are proposed, and each algorithm has their own advantages and disadvantages. The main purpose of this thesis is to solve problem that how to quickly finding content without enough information. We propose the fuzzy correlation used recommendation system concept to help user for retrieving the relevant articles. The article keywords are encoded as feature vectors to represent the article, we use the article keywords as feature vectors to represent the article, and find correlation between articles by fuzzy correlation coefficient. We use the article user viewed as the reference data to find the article has highly relevant for recommendation. Two steps in our proposed method to iterative recommend the more relevant article to use. First, compute and rank a correlation between the original article and the extension of the article through the simple fuzzy correlation coefficient. Second, we can get more information as reference data after the user click more articles. We use the result of last time recommendation as a new database, and then use multiple fuzzy correlation coefficient to find the articles which has highly relevant to the other article user viewed. We improve the accuracy of recommendation through the iterative recommendation step. We verify experimentally that this revised method, when used to text recommendation problem, outperforms than those methods designed for text recommendation problem.

第一章 緒論 1
1.1 前言 1
1.2 研究動機與目的 2
1.2.1 研究動機 2
1.2.2 研究目的 2
1.3 論文架構 3
第二章 相關背景知識 4
2.1 推薦系統 4
2.1.1 內容導向的推薦系統(Content-Base Commendation) 5
2.1.2 不分類的協同過濾推薦系統(Active Collaboration Filtering,ACF) 7
2.1.3 混合型推薦系統 9
2.1.4 其他類型的推薦系統 10
2.2 相關係數 10
2.3 糢糊集合 11
2.3.1 古典集合與糢糊集合 12
2.3.2 古典集合的定義與運算 12
2.3.3 糢糊集合的定義與運算 13
2.4 糢糊關聯分析 16
2.4.1 糢糊相關係數 16
2.4.2 多重糢糊相關係數 18
第三章 基於模糊關聯分析之論文推薦方法 23
3.1 架構與概念 23
3.2 架構流程 26
3.3 糢糊相關係數的應用 27
3.4 多重糢糊相關係數的應用 28
3.5 演算步驟 30
第四章 實驗 31
4.1 範例實驗 31
4.2 效能評估 45
4.2.1 評估方法 45
4.2.2 實驗結果 47
第5章 結論與未來研究的方向 49
5.1 結論 49
5.2 未來研究的方向 49
參考文獻 51
附錄-英文論文 53

圖目錄
圖1 以使用者輸入的關鍵字檢索出來的結果 23
圖2 擴大搜尋範圍找出與使用者第一次點選論文相關的論文 24
圖3 縮小範圍找出與使24用者第一、二次點選論文相關的論文 24
圖4 找出與閒錢點選過26的論文關聯性較高的其他論文，再次縮小範圍 25
圖5 使用者操作流程圖 26
圖6 precision/recall rate公式 46
 
表目錄
表1 6位學稱對模糊集合A和B的隸屬度 17
表2 Precision-recall 計算表格 46
表3 實驗結果比較 47

[1] T. W. Malone,K. R.Grant, S. A. Brobst, and M. D. Cohen. “Intelligent Information Sharing System.” Communications of the ACM, 30 ,5. PP.390-402(1987)
[2] Paul Resnick, Neophytos Iacovou, Mitesh Suchak, Peter Bergstrom, and John Riedl. “GroupLen: An Open Architecture for Collaborative Filtering of Netonews.” Proceedings of ACM 1994 Conference on Computer 30, 5. PP.390-402(1987)
[3] Shardanand U. and Maes, “Social information filtering: Algorithms for automating "word of mouth"”, Proceedings of CHI'95 -- Human Factors in Computing Systems, 210-217(1995)
[4] Badrul Sarwar, George Karypis, Joseph Konstan, and John Riedl. “Item-Based Collaborative Filtering Recommendation Algorithms”, WWW ACM, 1-5, 10, May, (2001) Hong Kong.
[5] Mark Balabanovic and Yoav Shoham. “Combining Contain-Based and Collaborative Recommendation.” Communication of the ACM, 5.(1997)
[6] Mark O'Connor and Jon Herlocker. “Clustering Items for Collaborative Filtering.” Department of Computer Science and Engineering, University of Minnesota.
[7] Jonathan L. Herlocker, Joseph A. Konstan, and John Riedl. Explaining Collaborative Filtering Recommendations. CSCW '00, December 2-6, (2000), Philadelphia
[8] Adomavicius, G., & Tuzhilin, A., “Extending recommender systems: A multidimensional approach,” Proceedings of the International Joint Conference on Artificial Intelligence (IJCAI-01), Workshop on Intelligent Techniques for Web Personalization, Seattle, (2001).
[9] Trotta, “Translating Strategy into Shareholder Value”, American Management Association,1st (2003)
[10] Bruke, R., “Knowledge-based Recommender Systems”, A. Kent(ed. ): Encyclopedia of Library and Information Systems, Vol. 69, Supplement 32,(2000)
[11] Joseph F. Healey, “Statistics: a tool for social research”. Thomson Wadsworth. 7th
[12] L. A. Zadeh, “Fuzzy sets”, Information and Control 8, 338-353 (1965).
[13] L. A. Zadeh, “Fuzzy sets as a basis for a theory of possibility”, Fuzzy Sets and Systems 1, 3-28 (1978).
[14] H. Bustince and P. Burillo, “Correlation of interval-valued intuitionistic fuzzy sets”, Fuzzy sets and Systems, 74, 237-244(1995)
[15] Ding-An Chiang和Nancy P. Lin, ”Correlation of fuzzy sets”, Fuzzy Sets and Systems, 102. 221-226(1999)
[16] Nancy P. Lin, ”Multiple Correlation Analysis of Fuzzy Sets”Proceedings of the International Conference on Computational Science, Melbourne, Australia. Part 4, pp.950-957 (2003.06)
[17] http://etds.ncl.edu.tw/
[18] M. Benkhalifa, A. Bensaid and A. Mouradi,  “Text categorization using the semi-supervised fuzzy  c-means algorithm,”  NAFIPS International Fuzzy Information Processing Society, pp.561- 565 (1999).
[19] Chun-Kai Chen, “A Study on Automatic Document Classification By Using Artificial Neural Network,” Master Degree Thesis, Department of Infoemation Engineering, Tamkang University, Taipei (1994). 
[20] Maron, Melvin E. "An Historical Note on the Origins of Probabilistic Indexing". Information Processing and Management 44: 971–972. (2008).
[21]  Michel W. Berry and Murray Browne, Understanding Search Engines: Mathematical Modeling and Text Retrival, USA,(1999)