標題: | 在行動社群網路中利用病毒式行銷推薦多個地點 Promoting Multi-location via Viral Marketing in Location-based Social Networks |
作者: | 温子毅 彭文志 莊仁輝 Wen, Zi-Yi Peng, Wen-Chih Chung, Jen-Hui 多媒體工程研究所 |
關鍵字: | 社群網路;資料探勘;地點推薦;行動預測;影響最大化問題;Location-based Social Network;Data Mining;mobility prediction;influence maximization problem;influence spread |
公開日期: | 2016 |
摘要: | 隨著社群網路的迅速成長,並且流行於提供人與人之間分享他們的活動、興趣等之類的平台。這些行動社群網路中 (像是Facebook、Foursquare) 允許使用者可以使用行動裝置透過在他們有興趣的地點上做打卡動作,藉由社群網路服務將打卡資訊分享給他們的朋友或者是家人。鑒於朋友之間的社群影響,許多興趣點 (電子地圖上的某個地點,用以標示該位置所代表的機構或景點) 已經吸引許多人去參觀。而地點推薦也頗受連鎖店的營銷策略所歡迎,許多業主希望更多人到他們的連鎖店進行消費行為。為了做到這一點,我們設定這些目標位置的廣告至個人,藉由社群網路的傳播,使得更多人可以得到個人有興趣的地點資訊。因此,我們制訂這方面的問題為多地點傳播的問題。也就是給定一組的目標位置資訊與一群傳播者,而目的就是藉由這群傳播者來影響最多的人可以到達我們目標位置進行活動。而我們所提出的多地點傳播模型,可以真實的反應在行動社群網路上多訊息的傳播的行為。基於現實生活中行動社群網路的實驗,我們可以證實我們的方法可以勝於其他先進的方法,表現出優異的效率與性能。 As the social networks has rapidly grown and become increasingly popular providing important platform for people to share their activities, interest and so on. The location-based social networking platform (e.g., Facebook, Foursquare) allow a user to check-in at a locations of interest with her mobile device, which reports visited locations to the LBSN. This information is then shared with other users who are socially related (e.g., friends, families). In view of the social influences of friends, recently, many POIs (Points of Interests) have explored check-in sharing to attract users to stay or visit. Location promotion is also popular with chain marketing, and many proprietors expect that many people will shop in their chain stores. To do this, we advertise the set of target locations to individuals, so that they can get the information about the set of target locations, which they are interested in. Therefore, we formulate this problem as a multi-location promotion problem. That is, given a set of target location and a set of seeds, the purpose is to maximize the number of influenced users. This paper proposes multi-location-aware propagation models to truly reflect the information propagation in LBSNs. Extensive experiments based on real LBSN datasets have demonstrated the superior effectiveness and performance of our proposals, which outperform the state-of-the-art algorithms. |
URI: | http://etd.lib.nctu.edu.tw/cdrfb3/record/nctu/#GT070256656 http://hdl.handle.net/11536/143102 |
Appears in Collections: | Thesis |