以社群網路為基礎的旅伴推薦系統

Abstract

臉書社群是目前世界影響最大的社群平台，根據官方統計截至2013年第三季已經有1.19億人次上線。隨著臉書使用者的增加其內容已經具有巨量資料的特性，而臉書使用者龐大的動態資訊隱藏各種有用資訊引起許多研究針對臉書使用參數，如按讚、社團參與次數等等作進一步的分析。 本研究以臉書社群為基礎，透過SCAN分群演算法的分析，讓使用者可就分群之結果了解周遭共同朋友的關係以及鏈結情況。此外，分群結果結合了人格特質模型，透過臉書參數，即可知道身邊朋友之人格特質。本研究並利用FQL擷取個人以及身邊朋友資訊做進一步的分析，使用者朋友越來越多時，利用SCAN分群能讓使用者在臉書人工分群的過程更有效率。最後，本系統並發展與旅遊資訊做整合用於自助旅行的功能以提供使用者針對特定旅遊行程找尋並推薦旅伴。Facebook is the most influential community platform in the world; according to the statistics officially released by Facebook, up to the third quarter of 2013, the number of active users has reached 1,190 millions. With the growth of users, Facebook has become a producer of Big data, and the vast amount of status updates posted by users, which conceal all kinds of useful information, have triggered many research projects focus on Facebook parameters such as the like count, the group count, etc. This research is based on Facebook communities. Through the analyses of the SCAN clustering algorithm, the users would be able to understand the relationships and connections between their mutual friends as stated in the results of grouping. Besides, the results of grouping have combined the Big Five personality model. By means of Facebook parameters, users can easily find out the personality traits of their friends. In this research, FQL was adopted to obtain the information of users themselves and friends around them for further analyses. When the number of friends is increasing, the SCAN clustering algorithm can improve the efficiency when it comes to grouping manually on Facebook. Lastly, this system has developed an integration with tourist information for the purpose of backpacking, so suitable travel companions will be recommended according to users' travel itineraries..