+ All Categories
Home > Data & Analytics > Seminar for undergrad

Seminar for undergrad

Date post: 15-Aug-2015
Category:
Upload: hoity-toity-shuvo
View: 119 times
Download: 0 times
Share this document with a friend
Popular Tags:
19
User Recommendations in Reciprocal and Bipartite Social Networks An online Dating Case Study Presented by: Md. Kamruzzaman
Transcript

User Recommendations in Reciprocal and

Bipartite Social NetworksAn online Dating Case Study

Presented by: Md. Kamruzzaman

Authors

Kang Zhao and Xi Wang, University of Iowa

Mo yu, Penn State University

Bo Gao, Beijing Jiaotong University

Social Network

• Facebook, LinkedIn,Twitter.• Billions of user.• Huge revenue opportunity.• Vast impact on social life.

Reciprocal Network

User

Bipartite Network

• Paper-Author network.• Audience-Movie network.

User Target

Reciprocal & Bipartite Network

• Online Dating Site, Matrimony Site.• Both reciprocal and bipartite network exist.

Goal

• Authors take online dating site for the research.• Better user recommendation is the ultimate goal.

User Recommendation

• Freinds of friends is the most popular method.• FoF is not possible for Reciprocal – bipartite network.• Existing system deals with explicit profile. Such as profession, age.• In this method, authors tried to find out a new way other than using

explicit profile.

Three Model System

• Baseline CF Model• Reciprocity-Only Model• Hybrid Model

Scenario

Baseline CF Model

Reciprocity Only Model

Hybrid Model

Calculation

• User, p• Target user, t

Using this formula given a success score for target user, the higher the score, the higher probability of recommendation.

Evaluation

• Two sets of metrics to evaluate• Initial contact

-IC precision@k-IC recall@k

• Reciprocal contact-RC precision@k-RC recall@k

Result

Result

Final Words

• According to experiment, Hybrid Model is more successful approach.

• Can be applied in university admission network, job hunting network.

• Use of sensitivity analysis can make HM more robust.

Thank You


Recommended