臺灣博碩士論文加值系統

個人化服務日漸普遍，發掘客戶偏好並提供適切的服務或產品推薦有益於提昇企業利潤。在網際網路環境中要進行個人化服務，網站大多需要建立詳細的個人輪廓(User Profile)，藉由收集使用者的網路瀏覽紀綠(Web Logs)，從中取得使用者偏好以提昇個人化推薦的準確度。
由於個人輪廓過去大多儲存於網站的伺服端，也因而產生了隱私權、一致性、及冷啟始(Cold-start)的問題，因此學者提出將個人輪廓儲存於客戶端的技術，稱之為通用個人輪廓(Universal User Profile)系統。它全面性的收集使用者瀏覽紀錄及提供單一的隱私資料授權入口而改善個人輪廓的問題。但網路使用者大多會瀏覽許多不同領域的網頁，瀏覽行為的多樣化及複雜性使得在客戶端發掘使用者偏好變得相當困難。
為克服以上問題，本研究建立一個結合本體論(Ontology)及通用個人輪廓的個人化推薦模式。首先以網路分類服務(Web Directory Service)作為本體論來辨識使用者在網路上的瀏覽行為，藉此發掘使用者的偏好；其次，利用網路習性探勘(Web Usage Mining)技術過濾多餘的瀏覽紀錄，增進個人化的準確度；最後，利用本體論的階層架構，從使用者偏好類別中發掘其潛在偏好，產生符合使用者特徵的通用個人輪廓。
本系統依照上述方法實作通用個人輪廓系統，並進行個人化推薦服務。實驗結果顯示使用通用個人輪廓的推薦準確率至少達到74%；加入網路習性探勘及本體論推論技術後，推薦準確率最高可達到90%。因此證明使用通用個人輪廓能達到良好的個人化服務，結合本體論更能提昇本系統的個人化效果。

Nowadays personalization services are getting popular than before. To discover customers preference, to provide needable offers or product recommendation are benefits in rising enterprise’s profits. To conduct personalized services in Internet environment, websites are acquired to establish intimate user’s profile for the most part. In addition, by collecting user’s web logs, user’s web usage behavior can be discovered and, therefore, the proper recommendation service can be achieved.
It is noticed, usually a user has too many user profiles stored in many web site’s. Because of it, the problems of privacy, consistency and cold-start has been brought into existence. Therefore, scholars have proposed the technology of storing user profiles in client-side, termed “Universal User Profile.” The technology wholly collect user’s web logs and provides unique portal of privacy data authorize to promote the problems of user profile. However most users are surf on various web pages, the diversification and complexity of user’s browsing behavior has turned the discovery of user’s preference more difficult.
To resolve the problems mentioned above, this research is aiming in building the personalized recommendation system by using ontology and universal user profile. Firstly, the website directory service is used as ontology to identify user’s browsing behaviors on Internet to discover user preferences. Secondly, redundant web logs are filtered out by using web usage mining, i.e., to enhance the accuracy of personalization. Finally, user’s potential preference is discovered by using the hierarchical property from user preference directory, that is, to bring out the universal user profile which match the characteristics of the user.
A design called Universal User Profile Recommendation System(UUPRS) has bulit to verify the research. In UUPRS the first step is to discover user’s preference by analyzing user’s log behavior. The second step is using Web Directory Service to build ontology tree. The ontology tree is used for creating user preference tree(UPT). In this research, three methods are designed to examine the effects of recommendation. These methods are weigthed sum method, session identification method and ontology inference method. The result of the experiment evidenced that the use of UUPRS can create the accuracy of 74%. With web usage mining and ontology, the recommendation accuracy to 90%. Therefore, this research proved that the use of UUPRS is able to promote personalized service and together with the ontology method is able to improve the positive effect of personalization.

摘　要 I
Abstract II
誌　謝 IV
第一章　緒論 10
1.1 研究背景 10
1.2 研究動機 12
1.3 研究目的 14
1.4 論文架構 15
1.5 研究範圍 16
第二章　文獻探討 17
2.1 個人化與個人輪廓(User Profile) 17
2.1.1 個人化(Personalization) 17
2.1.2 個人輪廓(User Profile) 20
2.1.3 個人輪廓與隱私權(Privacy) 22
2.2通用個人輪廓系統(Universal User Profile System) 24
2.2.1隱私偏好平台(Platform for Privacy Preferences, P3P) 26
2.2.2通用個人輪廓系統(Universal User Profile System) 29
2.3本體論 (Ontology) 33
2.3.1定義 33
2.3.2本體論與網路分類服務(Web Directory Service) 34
2.3.3本體論概念推論 37
2.4 網路習性探勘(Web Usage Mining) 39
2.4.1定義 40
2.4.2瀏覽序列識別(Session Identification) 41
2.5 小結 44
第三章　系統架構 45
3.1 系統概況 45
3.1.1使用者代理程式(User Agent) 46
3.1.2第三端伺服器(Third Party Server) 48
3.1.3網路服務提供者(Service Provider) 49
3.2網路分類檢索模組(Directory Search Module) 50
3.2.1前處理(Data Preprocess) 50
3.2.2分類檢索(Directory Search) 51
3.3 偏好推論模組( Preference Derive Module) 54
3.3.1權重累積法(Weighted Sum Method) 55
3.3.2序列識別法(Session Identification Method) 55
3.3.3本體論推論法(Ontology Inference Method) 58
3.4 偏好授權模組( Preference Authorized Module) 63
第四章　實驗 65
4.1 實驗架構及實施步驟 65
4.2 資料來源 66
4.2.1客戶端-使用者資料 66
4.2.2第三端網站-網站分類服務 68
4.3 分類搜尋模組處理程序 70
4.3.1網路瀏覽紀錄前處理 70
4.3.2網路分類搜尋 73
4.4 偏好推論模組處理程序 75
4.4.1累積權重法推薦列表 75
4.4.2序列識別法推薦列表 76
4.4.3個人偏好樹與本體論推論法推薦列表 79
4.5實驗評核指標、問卷設計及成效分析 83
4.5.1實驗評核指標及推薦問卷設計 83
4.5.2權重累積法實驗結果 84
4.5.3序列識別法實驗結果 86
4.5.4本體論推論法實驗結果 88
4.6 小結 89
第五章　結論 91
5.1結論與貢獻 91
5.2未來研究方向 92
附錄一、問卷範例 98
附錄二、權重累積法之推薦項目與使用者回饋表 101
附2.1 推薦列表喜好程度 101
附2.2 推薦項目誤差距離 105
附錄三、序列識別法之推薦項目與使用者回饋表 109
附3.1 推薦列表喜好程度 109
附3.2 推薦項目誤差距離 113
附錄四、本體論推論法之推薦項目與使用者回饋表 117
附4.1 推薦列表喜好程度 117
附4.2 推薦項目誤差距離 121

[1]邱永祥(2003)，運用類神經網路與資料探勘技術於網路教學課程推薦之研究，碩士論文，朝陽科技大學資訊管理系，台中。
[2]汪軒楷(2002)，策略式資料探勘在個人化推薦上之研究，碩士論文，真理大學管理科學研究所，台北。
[3]查士朗(2003)，網路時代個人資料管理系統之設計與實作，博士論文，國立台灣大學資訊管理系，台北。
[4]曾信誠(2004)，以本體論為基礎之使用者喜好萃取、隱私權控管與側寫建構，碩士論文，國立東華大學，花蓮。
[5]黎和欣(2001)，以單次匿名使用者為對象之概念導向資訊推薦機制，碩士論文，中華大學資訊工程系，新竹。
[6]廖婉菁(2002)，應用協同過濾機制於商品推薦之研究－以手機網站為例，碩士論文，中原大學資訊管理系，中壢。
[7]Arlein, Robert M., Ben Jai, Markus Jakobsson, Fabian Monrose, and Michael Reiter(2000), “Privacy-Preserving Global Customization,” The Second ACM conference on Electronic commerce, pp. 176–184.
[8]AT&T, Privacy Bird, Retrieve at: http://privacybird.com/.
[9]Balabanovic, Marko and Yoav Shoham(1997), “Fab: Content-Based, Collaborative Recommendation,” Communications of the ACM, Vol. 40, No. 3, pp. 66-72.
[10]Belkin, Nicholas. J. and W. Bruce Croft(1992), “Information Filtering and Information Retrieval: Two Sides of the Same Coin?,” Communications of the ACM, Vol. 35, No. 12, pp. 29-38.
[11]Cingil, Ibrahim, Asuman Dogac, and Ayca Azgin(2000), “A Broader Approach to Personalization,” Communications of the ACM, Vol. 43 No. 8, pp. 136–141.
[12]Cingil, Ibrahim(2002), “Supporting Global User Profiles through Trusted Authorities,” ACM SIGMOD Record, Vol. 31, pp. 11–17.
[13]Cooley, Robert, Bamshad Mobasher, and Jaideep Srivastava(1997), “Web Mining: Information and Pattern Discovery on the World Wide Web,” International Conference on Tools with Artificial Intelligence, pp. 558-567.
[14]Cooley, Robert, Bamshad Mobasher, and Jaideep Srivastava(1999), “Data Preparation for Mining World Wide Web Browsing Patterns,” journal of Knowledge and Information Systems, Vol. 1, No. 1, pp. 5-32.
[15]Dean, R. (1988), “Personalizing Your Web Site”, available at http://www.builder.com/business/personal.
[16]Fidel, Raya and Michael Crandall(1997), “Users'' Perception of the Performance of a Filtering System,” In Proceeding of the 20 Annual International ACM SIGIR Conference on Research and Development in Information Retrieval(SIGIR 1997) , pp. 198-205.
[17]Google Achieves Search Milestone with Immediate Access to More Than 6 Billion Items, Retrieve at: http://www.google.com.tw/press/pressrel/6billion.html.
[18]Hanani, Uri, Bracha Shapira, and Peretz Shoval(2001), “Information Filtering: Overview of Issues, Research and Systems,” User Modeling and User-Adapted Interaction(UMUAI) Vol. 11, Issue 3, pp. 203-259.
[19]Katzer, Jeffrey and Patricia T. Fletcher(1992), “The Information Environment of Managers,” Annual Review of Information Science and Technology, Vol. 27, pp. 227-63.
[20]Labrou, Yannis and Tim Finin(1999), “Yahoo! as an Ontology- Using Yahoo! Categories to Describe Documents,” Proceedings of the Eighth International Conference on Information and Knowledge Management(CKIM99), pp. 180-187.
[21]Microsoft, Microsoft Passport, Technical White Paper, March 2001.
[22]Middleton, Stuart E., Nigel R. Shadbolt, and David C. De Roure(2004), “Ontological User Profiling in Recommender Systems,” ACM Transactions on Information Systems, Vol. 22, Issue 1, pp. 54-88.
[23]Mittal, Banwari and Walfried M. Lassar(1987), “The Role of Personalization in Service Encounters,” Journal of Retailing, Vol. 72, No. 1, pp. 95-109.
[24]Mobasher, Bamshad, Honghua Dai, Tao Luo, and Miki Nakagawa(2001), “Effective Personalization Based on Association Rule Discovery from Web Usage Data,” Proceedings of the ACM Workshop on Web Information and Data Management (WIDM01), pp. 9-15.
[25]Noy, Natalya F. and Deborah L. McGuinness(2001), “Ontology Development 101: A Guide to Creating Your First Ontology,” Stanford Medical Informatics Technical Report ,SMI-2001-0880.
[26]Open Directory Project (ODP), retrieve at: http://dmoz.org/.
[27]Pfitzmann, Birgit (2003), “Privacy in Enterprise Identity Federation Policies for Liberty Single Sign-On,” Proceedings of the 3rd Workshop on Privacy Enhancing Technologies, pp. 26-28.
[28]Schafer, J. Ben., Joseph Konstan, and John Riedl(1999), “Recommender System in E-Commerce,” Proceedings of the first ACM conference on Electronic Commerce, pp. 158-166.
[29]Surprenant, C. F. and M. R. Solomon(1987), “Predictability and Personalization in the Service Encounter,” Journal of Marketing, Vol. 51, pp. 86-89.
[30]The World Wide Web Consortium – Platform for Privacy Preferences (P3P), W3C Resource, Retrieve at: http://www.w3.org/P3P/.
[31]Wang, Huaiqing, Matthew K. O. Lee, and Chen Wang(1998), “Consumer Privacy Concerns about Marketing,” Communications of the ACM, Vol. 41, Issue 3, pp. 63-70.
[32]Web-Ontology(WebOnt) Working Group, Retrieve at: http://www.w3.org/2001/sw/WebOnt/.
[33]WordNet, Retrieve at: http://www.cogsci.princeton.edu/~wn/
[34]Yahoo! Directory, Retrieve at: http://dir.yahoo.com/.