Abstract—To reduce the difficulty of personalized recommendations, the traditional network-based method constructed bipartite networks with stronger links (higher ratings). However, weaker links and link weights were almost ignored. Although the existing method effectively mined users’ preferences, it was impossible to catch users’ disgusts. Therefore, this paper proposed a novel method to effectively discover users’ preferences and disgusts. Experimental results on the MovieLens dataset demonstrated that the proposed method was much more superior to the baseline method under the diversity index.
Index Terms—Personalized recommendations, weighted bipartite network, users’ preferences, users’ disgusts, diversity.
Jing Wang, Fengjing Shao, Shunyao Wu, Rencheng Sun, and Ran Li are with Qingdao University, Shandong, China (e-mail: sfj@qdu.edu.cn).
[PDF]
Cite:Jing Wang, Fengjing Shao, Shunyao Wu, Rencheng Sun, and Ran Li, "Weighted Bipartite Network Projection for Personalized Recommendations," Journal of Advances in Computer Networks vol. 4, no. 1, pp. 64-69, 2016.
