陳銘憲Chen, Ming-Syan臺灣大學:電機工程學研究所蔡尚安Tsai, Shang-AnShang-AnTsai2010-07-012018-07-062010-07-012018-07-062009U0001-0608200915361200http://ntur.lib.ntu.edu.tw//handle/246246/188121適地性服務是指根據使用者所在地理位置提供服務的應用。不同種類的適地性服務因為無線通訊技術、可攜性裝置和位置定位如全球定位系統的快速成長，正蓬勃發展。近年來，有更多的適地性服務針對手機用戶被開發出來，並更進一步考慮使用者資訊如興趣、購買行為等。對一群使用者如朋友來說，有很大的機會他們會願意參加可一起進行的社交活動，如一起購物。因此，適地性服務在提供推薦的時候，必須能夠考量所有使用者的情況。但在傳統的適地性服務，大多針對單一使用者提供服務，也就是說僅考慮單一使用者地點資訊來做推薦。了要提供一群使用者適地性服務，我們提出了一個位置感知top-k時空間事件查詢的系統。具體來說，我們利用使用者群之間的社交關係，提供一個群排名的方法。在我們的設計中，我們根據使用者群內每位使用者的位置還有時間限制，找到一些候選事件。接著，依照每位使用者的資訊和位置，對這些候選事件評分、得到個人的事件排名。 而後，利用使用者群間的社交關係、個人事件排名、以及使用者的位置來找出此使用者群的事件排名。最後，排名前k名的事件將會推薦給使用者。實驗結果顯示，我們提出的系統運作良好，並且在移動式的環境、高差異的社交影響力下，對非大數量的使用者群能獲得高的群體滿意度。A Location-based Service, which is abbreviated as LBS, is defined as an application that serves a user based on his/her physical locations. Various kinds of LBS are on the rise and flourishing because of the rapid development of wireless technologies, mobile devices, and position systems. Through these technologies and devices, the LBSs are developed to assist people to search, browse or interact with items physically around them. Recently, more LBSs are developed to serve mobile users, which further consider user profiles such as the user’s interest and their shopping behavior. When a group of users would like to engage in social activities together, like shopping, an LBS should consider all the user preferences in recommending a list of suitable options. However, the traditional LBSs are usually designed to serve a single user. That is, a recommendation is made to a single user based only on his profile and location. herefore, to provide LBSs for multiple users, we propose a location-aware top-k spatial-temporal event query system. Specifically, we propose a new group ranking function that considers the social relationships between the users in a group. In our design, we initially find a list of events based on the time constraints and locations of all mobile participants. Next, we grade the events according to the location and the profile of each user. Then, our algorithm ranks the candidate events based on each personal ranked list, the social network and the locations of the users. Finally, top-k events are returned. The experiment results show that the proposed system works well and receives high group satisfactions in a mobile environment when the group size is not very big, and the social impacts of people in the same group have high divergence.口試委員會審定書 #cknowledgements i要 iiBSTRACT iiiONTENTS ivIST OF FIGURES viIST OF TABLES vii Introdution 1 Related Work 5.1 Context-Aware Recommendation Systems 5.2 Mobile Social Networks 6.3 Recommendation Systems with Social Network Analysis 6.4 Multiple Criteria Group Decision-Making 7 System Overview 9.1 Server Side 10.2 Mobile Client 11.3 Procedures in the System 12.3.1 An Example and the System Work Flow 12.3.2 Discussion 18 Processing Spatial-temporal Event Queries by Utilizing Social Networks 19.1 Alternatives Generation 21.2 Personal Ranked List Generation 22.2.1 Criterion Evaluation and Normalization 22.2.2 Criterion Weighting and Final Score Calculation 24.3 Public Consensus Ranked List Generation 25.3.1 Social Impact Computation - UserRank 26.3.2 Group Decision Making – BordaRank 29.4 Social Network Updating 31 Experiment Results 32.1 Experiment Settings 32.1.1 The Personal Satisfaction and the Group Satisfaction 32.1.2 Attending Simulation 33.2 Performance Study with Synthetic Data 34.2.1 The Group Size 34.2.2 The Query Time and the Social Impact 35.2.3 The Mobility 37 Conclusion 39EFERENCE 411382877 bytesapplication/pdfen-US適地性服務推薦系統社交網路分析群體決策行動社交網路location-based servicesrecommendation systemsocial networks analysisgroup decision makingmobile social networkthesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/188121/1/ntu-98-R96921021-1.pdf