透過您的圖書館登入
IP:3.17.175.182
  • 學位論文

社群中使用者對項目的鏈結預測

Link Prediction for Social User-Item Networks

指導教授 : 張正尚

摘要


推薦系統是一個用來幫助使用者找到他們有興趣物品的熱門工具,可見於許多電子商務網站上,如:亞馬遜(線上購物網站)、Netflix(影片租借平台)。此系統根據使用者過去的購買紀錄,推測使用者可能的興趣,進而達成個人化推薦之目的。 本研究以DBLP資料庫(一個文獻搜索資料庫)為基礎進行鏈結預測之研究。目標是從發表者過去所參與的會議,來推測他未來可能會參與的會議。亦即,預測發表者在未來可能會何處發表論文。 協同過濾(collaborative filtering)是推薦系統中廣泛被應用的技術。但許多文獻皆指出此技術有網路稀疏(sparsity)的嚴重限制。舉例而言,在一個推薦系統網路中,每個人曾經購買的物品可能僅僅幾十個項目,相對於系統中數以萬計的項目數來說,可說是微乎其微,進而導致此技術可能無法產生有效的推薦。 因此,本研究以隨機漫步(random walk)的方法來解決此問題:透過馬可夫鏈(Markov chain)中的轉移機率,來計算發表者將在某個會議上發表論文的機率。此作法所傳達的概念和協同過濾是類似的,同時亦解決了網路稀疏的問題。此外,本研究除了在二分圖(bipartite graph)上隨機漫步之外,還加入了作者間的友誼關係,此訊息可使預測更加精準。 研究發現,調整隨機漫步者的前進機率後,適當的參數加上朋友關係可以使得預測更加準確。同時,時間因素也對於推薦系統有重要的影響。此外,當推薦系統推薦較多的清單量給使用者時,選擇較小的前進機率更能增加推薦的準確度。

關鍵字

網路科學 二分圖 推薦 預測

並列摘要


Recommendation is the most popular tool to help users find the new items they are interested in. We study the link prediction problem on the author-conference network of DBLP data set, and we would like to predict what conferences the author will publish in. Collaborative filtering is the most common method to suggest items for users. However, the limitation of this approach is the sparsity problem. As a result, we perform the random walk on the graph to calculate the transition probability for predicting. We consider not only the bipartite graph but also the relationship of these authors, so we perform the random walk on this union of two graphs. Experimental results show it can predict more precisely when choosing the appropriate parameters in our algorithm, and it is useful with the information of the friendship.

參考文獻


[4] G. Linden, B. Smith, and J. York, “Amazon. com recommendations: Item-to-item collaborative filtering,” Internet Computing, IEEE, vol. 7, no. 1, pp. 76–80, 2003.
[7] D. Goldberg, D. Nichols, B. M. Oki, and D. Terry, “Using collaborative filtering to weave an information tapestry,” Communications of the ACM, vol. 35, no. 12, pp. 61–70, 1992.
[8] D. M. Dunlavy, T. G. Kolda, and E. Acar, “Temporal link prediction using matrix and tensor factorizations,” ACM Transactions on Knowledge Discovery from Data (TKDD), vol. 5, no. 2, p. 10, 2011.
[12] S. Brin and L. Page, “The anatomy of a large-scale hypertextual web search engine,” Computer networks and ISDN systems, vol. 30, no. 1, pp. 107–117, 1998.
[13] S. Baluja, R. Seth, D. Sivakumar, Y. Jing, J. Yagnik, S. Kumar, D. Ravichandran, and M. Aly, “Video suggestion and discovery for youtube: taking random walks through the view graph,” in Proceedings of the 17th international conference on World Wide Web. ACM, 2008, pp. 895–904.

延伸閱讀