網際網路的快速發展使得推薦系統廣泛被運用在網站上。推薦系統可以幫助使用者減少搜尋項目的時間並提供優選項目給使用者。過去有關推薦系統的研究中，協同過濾式推薦已廣泛且成功地被應用於網路推薦上，隨著近年社群網路的發展，現在許多學者亦透過使用者彼此間的信任關係做為協同過濾(Collaborative Filtering)推薦之研究，並稱之為以信任為基礎的推薦系統(Trust-based Recommender System: TBRS)，此系統中最重要的一項研究議題即如何定義最佳的信任初始值。
過去研究有關信任初始值的設定通常為隨機或直接設定，然而這些初始值均未考量使用者間的正向或負向關係，為了能找出使用者間的正負向關係，本研究運用社群間的使用者評論作為分析使用者間的關係，運用TF-IDF算法計算每個評論的特徵向量，再將評論之特徵向量輸入支持向量機(Support Vector Machine)進行信任模型的訓練，最後結合信任模型(Trust Model)與信任推演模型(Trust Propagation Model)建立個人的信任網路(Trust Network)。
本研究提出一個以網路社群評論信任值為基礎之協同過濾式推薦系統，本研究改善過去信任初始值之設定方法，經實驗結果顯示，本研究之成果優於其它以信任為基礎之推薦方法。

Due to the repaid growth of Internet, the recommender systems are widely used as the network service. Recommender system can help user reducing the search cost and providing a list of suitable items for the user. In the past study, Collaborative Filtering Recommendation (CF) has been widely applied and successfully used in the Internet. Now, with the popularity of social network, many researchers have proposed using trust between users for collaborative filtering recommendation. This kind of approach we also call it as the Trust-based Recommender System (TBRS). And, one of the important issues about TBRS is how to find and define the optimal trust values.
In the past study, the initial trust value is usually set by a random number or directly set by a certain value. The initial trust values shall represent positive or negative relationship between users. However, the positive/negative relationship of initial trust values did not considered in the past studies. In order to find out the positive/negative relationship, this research analyzed the users’ comment of social network. The features of these comments are analyzed and derived by using Term Frequency-Inverse Documents Frequency (TF-IDF) method. To obtain the training model for finding the positive/negative relationship, these features are then classified by using Support Vector Machine (SVM). The obtained Trust Model and the Trust Propagation Model are then integrated to generate the user’s trust network model.
This research has proposed a trust based recommender system by analyzing user comments of social network. The proposed method improves the finding of initial trust value. From the experiments results, it proves that the proposed method has better prediction outcome than other methods.

目 錄
中文摘要 I
英文摘要 II
誌謝 III
目錄 Ⅳ
圖目錄 Ⅵ
表目錄 Ⅶ
壹、 導論 1
貳、 文獻探討 6
2.1 推薦系統 6
2.1.1 內容導向式推薦系統 7
2.1.2 協同過濾式推薦系統 8
2.1.3 協同過濾式推薦系統的預測 9
2.2 以信任為基礎的推薦系統 10
2.2.1 信任初始值 13
2.2.2 信任值修正 15
2.2.3 信任推演 15
2.3 Support Vector Machine(SVM) 19
2.3.1 SVM算法 21
2.4 TF-IDF 23
2.4.1 TF-IDF公式 24
參、 研究方法 26
3.1 研究流程 27
3.2 使用TF-IDF擷取評論特徵 29
3.3 使用SVM評論分類 30
3.4 信任值計算 32
3.5 以信任為基礎的評分預測 34
3.6 使用者回饋機制 35
肆、 實驗結果 37
4.1 資料蒐集 38
4.2 信任初始值的運算 40
4.2.1 SVM分類準確度評估 40
4.3 使用者信任網路的建立 41
4.3.1 以信任為基礎的協同過濾預測 42
4.3.2 預測準確度評估 44
伍、 實驗結果及未來研究 47
5.1 研究結果 47
5.2 未來研究 48
參考文獻 50

圖目錄
圖 1: 信任網路 13
圖 2: 直接推演 17
圖 3: 共同信任 17
圖 4: 信任轉換 18
圖 5: 耦合信任 18
圖 6: 支持向量 22
圖 7: 透過網路上使用者評論擷取信任值 27
圖 8: 研究流程 28
圖 9: 計算特徵向量之步驟 30
圖 10: 透過Sigmoid方法轉換SVM輸出結果 32
圖11: 使用者對產品的評分與評論 39
圖12: 其他使用者討論與評論 39
圖13: 使用者信任矩陣 42
圖14: 使用者評分資料 43
圖15: 使用者信任值修正 43
圖16: 協同過濾式預測結果 44
圖17: 加入使用者回饋機制後的預測結果 45

表目錄
表 一: 推薦技術分類表 7
表 二: 以信任為基礎的推薦系統研究整理 11
表 三: 回應率與準確率 41
表 四: 不同信任模型的MAE 46

[1]陳寶旭，”運用協同過濾推薦機制於搭售商品之推薦”，朝陽科技大學資訊管理系碩士論文，2006。
[2]廖品妍，”以顯性評價為主之相似性推薦”，朝陽科技大學資訊管理系碩士論文，2010。
[3]蔡美慧，”以偏好導向為相似性量測的協同過濾式推薦”， 朝陽科技大學資訊管理系碩士論文，2010。
[4]許榮望，”評論語意自動化分析與顧客滿意度應用之研究-以中國旅遊網站為例”，朝陽科技大學資訊管理系博士論文，2013。
[5]遊和正、黃挺豪、陳信希，”領域相關詞彙極性分析及文件情緒分類之研究”，中文計算語言學期刊，第17卷第四期，第33-48頁，2012。
[6]侯彥旭，”以平價理論為基礎的電子本文分類”，中原大學資訊管理學系碩士論文，2010。
[7][鍾任明、李維平、吳澤民，”運用文字探勘於日內股價漲跌趨勢預測之研究”，中華管理評論，第十卷第一期，2007。
[8]黃河銓、李奕縉、黃昱凱，” 整合信任網路與回饋機制之個人化餐廳推薦系統”，資訊與管理科學，5(2)，65-78，2012。
[9]Alfarez Abdul-Rahman and Stephen Hailes, “Supporting Trust in Virtual Communities,” Proceedings of the 33rd Annual Hawaii International Conference on System Sciences, Hawaii, pp. 1-9, 2000.
[10]Gediminas Adomavicius, Alexander Tuzhilin, “Towards the Next Generation of Recommender Systems,” A Survey of the State-of-the-Art and Possible Extensions, IEEE Transactions on Knowledge and Data Engineering, vol. 17, no. 6, pp. 734-749, 2005.
[11]Bernhard E. Boser, Isabelle M. Guyon and Vladimir N. Vapnik, “A Training Algorithm for Optimal Margin Classifiers,” the 5th Annual ACM Workshop on COLT, New York, pp. 144-152, 1992.
[12]Federico Chesani, “Recommendation Systems.” Corso di laurea in Ingegneria Informatica, 2002.
[13]Nello Christianini and John Shawe-Taylor, “An Introduction to Support Vector Machines and Other Kernel-based Learning Methods,” Cambridge, U.K.: Cambridge Univ. Press, vol. 18, no. 6, pp 687-689, 2000.
[14]Iván Cantador, Alejandro Bellogín, David Vallet, 2010 “Content-based Recommendation in Social Tagging Systems,” Proceedings of the 4th ACM conference on Recommender systems, USA, pp. 237-240, 2010.
[15]Susan Dumais, John Platt, David Heckerman, “Inductive Learning Algorithms and Representations for Text Categorization,” Proceedings of the 7th International Conference on Information and Knowledge Management, pp. 148-155, 1998.
[16]Abhinandan Das, Mayur Datar and Ashutosh Garg, “Google News Personalization: Scalable Online Collaborative Filtering,” Proceeding of 16th International Conference on World Wide Web, Canada, pp. 271-280, 2007.
[17]Jennifer Golbeck, “Generating Predictive Movie Recommendations from Trust in Social Network, ” Proceedings of The 4th International Conference on Trust Management Lecture Notes in Computer Science, vol. 3986, pp. 93-104, 2006.
[18]R. Guha , Ravi Kumar, “Propagation of Trust and Distrust,” In proceeding of the 13th International Conference on World Wild Web, New York, pp. 403-412, 2004.
[19]Omar Hasan, Lionel Brunie and Jean-Marc Pierson, “Evaluation of the Iterative Multiplication Strategy for Trust Propagation in Pervasive Environments,” Proceedings of the 2009 International Conference on Pervasive services, USA, pp. 49-54, 2009.
[20]Chein-Shung Hwang and Yu-Pin Chen, “Using Trust in Collaborative Filtering Recommendation,” New Trends in Applied Artificial Intelligence, Vol. 4570, pp. 1052-1060, 2007.
[21]Audun Josan, Stephen Marsh and Simon Pope, “Exploring Different Type of Trust Propagation, ” Proceeding of the 4th International Conference on iTrust, Ltaly, pp. 179-192, 2006.
[22]Mohsen Jamali, Martin Ester, “A Matrix Factorization Technique with Trust Propagation for Recommendation in Social Networks,” Proceeding of the 4th ACM Conference on Recommender Systems, Barcelona, pp. 135-142. 2010.
[23]Banage T. G. S. Kumara, Incheon Paik, Hiroki Ohashi, Yuichi Yaguchi and Wuhui Chen, “Context-Aware Filtering and Visualization of Web Service Clusters,” Proceeding of IEEE International Conference on Web Services (ICWS), Alaska, pp. 89-96, 2014.
[24]Young Ae Kim, Hee Seok Song, “Strategies for Predicting Local Trust based on Trust Propagation in Social Networks,” Knowledge-Based Systems, vol. 24, no. 8, pp. 1360-371, 2011.
[25]Neal Lathia, Stephen Hailes and Licia Capra, "Trust-Based Collaborative Filtering," Trust Management II: Proceedings of IFIPTM 2008: Joint iTrust and PST Conferences on Privacy, Trust Management and Security, June 18-20, 2008, Trondheim, Norway, pp. 119, 2008.
[26]Nan Li, Desheng Dash Wu, “Using text mining and sentiment analysis for online forums hotspot detection and forecast,“ Decision Support Systems vol. 48, no. 2, pp. 354–368, 2010.
[27]Paolo Massa, Paolo Avesani, “Trust-aware Recommender Systems.” Proceeding of the ACM Conference on Recommender Systems, Minnesota, USA, pp. 17-24. 2007.
[28]Samaneh Moghaddam, Mohsen Jamali, Martin Ester, “Feedback Trust: Using Feedback Effects in Trust-based Recommendation Systems,” Proceedings of the 3th ACM Conference on Recommender Systems, New York, USA. pp. 269-272, 2009.
[29]Raymond J. Mooney and Loriene Roy, “ContentBased Book RecommendingUsing Learning for Text Categorization,” Proceedings of the 4th ACM Conference on Recommender Systems, Barcelona, Spain, pp. 26-30, 1999.
[30]John O’Donovan, Barry Smyth, “Trust in Recommender Systems,” IUI ’05 Proceedings of the 10th International Conference on Intelligent User Interfaces, New San Diego, pp. 167-174, 2005.
[31]Incheon Paik and Hiroshi Mizugai, “Recommendation System Using Weighted TF-IDF and Naïve Bayes Classifiers on RSS Contents,” Journal of Advanced Computational Intelligence and Intelligent Informatics, vol. 14, No. 6, 2010.
[32]John C. Platt, “Probabilistic Outputs for Support Vector Machines and Comparisons to Regularized Likelihood Methods,” Advances in large margin classifiers, 1999.
[33]Michael J. Pazzani and Daniel Billsus, 2007, “Content-Based Recommendation Systems,” Springer-Verlag Berlin Heidelberg, LNCS 4321, pp. 325 – 341, 2007.
[34]Catarina Silva, Bernardete Ribeiro, “On text-based mining with active learning and background knowledge using SVM,” Soft Computing, vol. 11, no. 6, pp. 519-530, 2006.
[35]Georges Siolas, Florence d’Alche-Buc, “Support Vector Machine based on a Semantic Kernel for Text Categorization.” Proceedings of the IEEE-INNS-ENNS International Joint Conference on Neural Network, Como, vol. 5, pp. 205-209, 2000.
[36]Pooja Vashisth, Deepak Chandoliya, Bipin Kumar Yadav, Punam Bedi, “Trust enabled Argumentation Based Recommender System,” Proceeding of 12th International Conference on Intelligent Systems Design and Applications (ISDA), Kochi, Japan, pp. 137-142, 2012.
[37]Vladimir Vapnik and Corinna Cortes, “Support vector networks,” Machine Learning, vol. 20, no. 3, pp. 273–293, 1995.
[38]Jianshu Weng, Chunyan Miao and Angela Goh, “Improving Collaborative Filtering with Trust-based Metrics,” Proceedings of the 2006 ACM symposium on Applied computing (SAC '06), pp. 1860-1864, 2006.
[39]Sheng-Han, Yang, Incheon Paik, Li-Hua Li, “Improving Trust Based Recommender System By Mining User Comments on the Web,” The 2015 World Congress on Information Technology Applications and Services, Jeju, Korea, 2015.
[40]Cai-Nicolas Ziegler and Georg Lausen, “Spreading Activation Models for Trust Propagation,” Proceedings of 2004 IEEE International Conference on e-Technology, e-Commerce and e-Service, pp. 83-97, 2004.
[41]http://epinion.com (2014)
[42]http://amazon.com (2014)
[43]http://svmlight.joachims.org/