Presentation is loading. Please wait.

Presentation is loading. Please wait.

Community-based User Recommendation in Uni-Directional Social Networks

Similar presentations


Presentation on theme: "Community-based User Recommendation in Uni-Directional Social Networks"β€” Presentation transcript:

1 Community-based User Recommendation in Uni-Directional Social Networks
Gang Zhao, Mong Li Lee, Wynne Hsu, Wei Chen, Haoji Hu School of Computing, National University of Singapore

2 Contents

3 Purpose Design an user recommendation system in Twitter-style social network Find a set of users whom a target user is likely to follow

4 Challenges Tweet comments are typically short and noisy
Data is very sparse

5 Proposed Solution Forming communities to reduce data sparsity
Applying matrix factorization on each communities

6 Twitter-style Social Network

7 Discover Communities Framework (1)
U is the set of all users F is the set of followers G is the set of followees 𝑑 𝑓 is the list of followees of user u 𝑑 𝑔 is the list of followers of user u

8 Discover Communities Framework (2)
Choose number of topics Apply Latent Dirichlet Allocation (LDA) to determine the topic distribution of users For each topic z, form a community c:

9 Recommend Followees (1)
For each community c, construct matrix M with size |c.F| x |c.G| Apply Implicit Feedback-Matrix Factorization (IF-MF) Obtained matrix and

10 Recommend Followees (2)
Row vectors associate with followers Column vectors associate with followees

11 Datasets

12 Evaluation Metrics

13 Experiments (1)

14 Experiments (2)

15 Q & A


Download ppt "Community-based User Recommendation in Uni-Directional Social Networks"

Similar presentations


Ads by Google