適地性社群網絡上的使用者影響力探勘

近年來，隨著 Foursquare、Yelp、Geolife 及 Flickr 等等適地性社群網絡 (location-based social network, LBSN)服務的蓬勃發展，累積產生的 LBSN 記錄也為旅遊規劃及景點資訊推薦相關研究開創了新的方向。現有的研究大多著重於透過群眾力量找出熱門景點(Point-of-interest, POI)，並依使用者需求產生推薦。舉例來說，從個人的位置歷史提取其行為偏好，用以為查詢區域中的 POI 給定推薦度分數。然而，這類研究忽略了隱藏在 LBSN 中的「社群關係」，亦即除了自身興趣外還有來自身邊他人的影響力。假設我們得到了一個推薦拜訪地點的建議，來自朋友的好評推薦應該會比陌生群眾的建議更容易被接受。所以在本文中，我們提出了一個全新的「以社群影響力為基礎的使用者推薦機制」（social influence-based user recommender , SIR），不同於以往的研究以地點為主，它旨在從可靠的用戶（親密的朋友或旅遊專家等）發掘潛在價值。明確地說，我們的 SIR 機制可以推斷出 LBSN 中有影響力的使用者。我們結合了虛擬社群、現實生活中的移動行動與時間效應之間的相互作用，來推斷任意用戶組之間的社群影響力。此外，我們以熱量的擴散作用來模擬影響力在使用者之間的傳遞。第三，我們設計了一個動態的融合架構，整合發掘出來的特徵為一個反應受使用者影響概率的分數。最後，SIR 機制針對個人提供個性化的 k 個使用者的推薦。為了驗證推薦結果，我們使用真實數據集（Flickr 資料和 Gowalla 資料）進行實驗，實驗結果顯示我們的 SIR 架構的在預測的準確度和推薦的可靠度皆優於過去的推薦機制。

關鍵字

適地性社群網絡；資料探勘

並列摘要

Recently, with the advent of location-based social networking services (LBSNs), e.g., Foursquare, Yelp, Geolife and Flickr, travel planing and location-aware information recommendation based on LBSN have attracted research attentions, such as Point-of-interest (POI) recommendation or travel routes recommendation. Most of the existing works mainly address on mining POIs by crowd power in LBSN and generate the recommendation on demand. For example, extracting personal preferences from individual’s location history to score the POIs in query region. However, they ignore the impact of social relations hidden in LBSN, i.e., the social influence from friends. We suppose that the location-aware recommendation with favorable comments from friends should be more reliable than that from unfamiliar crowd.In this paper, we propose a new social influence-based user recommender framework (SIR) which differs from prior works by aiming to discover the potential value addition from reliable users (i.e. close friends and travel experts). Explicitly, our SIR framework is able to infer influential users from LBSN. We claim to capture the interaction among virtual community, physical mobility activities and time effects to infer the social influence between user pairs. Furthetmore, we intend to model the propagation of influence in terms of diffusion-based mechanism. Third, we design a dynamic fusion framework to integrate the features mined into an united follow probability score. Finally, our framework provides personalized top-k user recommendation for individuals. To evaluate the recommendation results, we conduct extensive experiments on real datasets (i.e., Flicker dataset and Gowalla datasets). The experimental results shows the performance of our SIR framework is better than a state-of-the-art user recommendation mechanisms in terms of accuracy and reliability.

並列關鍵字

location-based social network ； data mining

參考文獻

[2] M. De Choudhury, M. Feldman, S. Amer-Yahia, N. Golbandi, R. Lempel, and C. Yu, “Automatic construction of travel itineraries using social

generating travel routes from geo-tagged photos for trip planning,” in

2010, pp. 143–152.

international conference on Web search and data mining. ACM, 2012,

conference on World wide web. ACM, 2009, pp. 791–800.

國際替代計量

適地性社群網絡上的使用者影響力探勘

全文下載

主題瀏覽