Facebook’s public social interaction utilization to assist recommendation across system domain

Social media is most prominent internet transition for this decade and Facebook holds its largest share. Facebook has been utilized by researchers from different perspectives e.g. opinion mining, user mood swing pattern, influential person identification etc. whereas recently Facebook’s social interactions were used for recommendation purposes. Although social interactions assisted recommendation, these interactions forced algorithm to work inside Facebook’s ecosystem i.e. recommending items existing inside Facebook to Facebook users and these interactions were private in nature, requiring explicit permission from user before algorithm execution. This study utilize Facebook’s public social interactions to recommend items across system domain i.e. recommending items to users existing outside Facebook. For this purpose we propose an algorithm that first identify items on Facebook’s public pages, gather social interactions related to them, generate a rank list and finally recommend it to external users. As an experimental case study, “whatmobile.pk” Facebook’s public social page was scanned for items and respective social interactions. These items were then compared with “fan” attribute of items existing on GSMARENA.com website in order to show rank similarity. 299 total items were found common between Facebook’s public page and GSMARENA website. Items were ranked according to social interactions and “fans” quantity. Then a positive spearman correlation of 0.547 was found which was improved to 0.660 by excluding 22 mobile phones.

[1]  Tsvi Kuflik,et al.  Domain ranking for cross domain collaborative filtering , 2012, UMAP.

[2]  Paolo Cremonesi,et al.  Cross-Domain Recommender Systems , 2011, 2011 IEEE 11th International Conference on Data Mining Workshops.

[3]  Qian Xu,et al.  Social-behavior Transfer Learning for Recommendation Systems , 2011 .

[4]  Lior Rokach,et al.  Facebook single and cross domain data for recommendation systems , 2013, User Modeling and User-Adapted Interaction.

[5]  Bin Li,et al.  Cross-Domain Collaborative Filtering: A Brief Survey , 2011, 2011 IEEE 23rd International Conference on Tools with Artificial Intelligence.

[6]  L. Myers,et al.  Spearman Correlation Coefficients, Differences between , 2004 .

[7]  John Riedl,et al.  Ganging up on Information Overload , 1998, Computer.

[8]  Marshall McLuhan,et al.  The Global Village: Transformations in World Life and Media in the 21st Century , 1989 .

[9]  Lior Rokach,et al.  Introduction to Recommender Systems Handbook , 2011, Recommender Systems Handbook.

[10]  Dan Zhang,et al.  Multiple Instance Transfer Learning , 2009, 2009 IEEE International Conference on Data Mining Workshops.

[11]  Chang-Hyun Jin,et al.  The role of Facebook users’ self-systems in generating social relationships and social capital effects , 2015, New Media Soc..

[12]  Guy Shani,et al.  TALMUD: transfer learning for multiple domains , 2012, CIKM.