Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
Search
Beyond social graphs: mining patterns underlying social interactions
Baldominos A., Calle J., Cuadra D. Pattern Analysis & Applications20 (1):269-285,2017.Type:Article
Date Reviewed: Jul 26 2017

Reliable recommendation systems are the holy grail of Internet research these days. They suggest new links to social network users. Because the attractiveness of a social network depends on the number of nodes and links composing it, the more nodes and links, the more valuable a social network is. Hence the importance of these systems. The authors of this paper build a recommendation system using off-the-shelf components, namely Cognos, a social media analytics solution by IBM, and data extracted from Facebook with the Facebook Graph application programming interface (API). What is new here is how authors combine these results and enhance their power.

The paper starts by explaining the different parts of a recommendation system: community detection; social classification, or assigning weights to nodes and links; social recommendation through collaborative and content-based filtering; and graph visualization via graphical or matrix representations of nodes and links. It then describes some algorithms currently in use. The paper goes on, presenting the authors’ own recommendation system, which, as said before, takes data from existing applications but combines them in a novel way that allows for the creation of far richer graphs. First, data representing Facebook users is extracted from Facebook using the Facebook Graph API, and then the MapReduce paradigm is applied to this data in order to center data on each single user; the authors call these resulting graphs “ego graphs.” Social information is then grouped into social communities by way of the Louvain method, and then patterns are found by computing degrees of intersection among all community features; at this stage the paper is still dealing with single ego graphs. Next, single ego graphs are aggregated by repeatedly applying these same patterns to every node in turn. Finally, the recommendation engine is built by aggregating ego graphs at the community level. For most of these steps the paper gives pseudocode for the Cognos software; mathematical properties of the algorithms are also described.

The paper then presents a validation and evaluation section for the whole system. First, the experimental setup is described with a table presenting the overall properties of the links considered. Next, a subjective (qualitative) evaluation asks people how much they agree with the results from the recommender system. Finally, an objective (quantitative) evaluation is performed by taking system nodes, removing some of their links, letting the system again recommend links for them, and evaluating how many of the original links the system recommends again. The paper ends with references and pointers to future work; this of course can be a hint for other teams working in the same field. In my opinion, the most striking feature of the paper is how the authors, starting from widely available components, build something entirely new and much more powerful than the initial components’ features. Although developed in a research environment with an academic target, this paper is for anyone interested in gaining a deeper understanding of the forces behind social networks.

Reviewer:  Andrea Paramithiotti Review #: CR145445 (1710-0673)
Bookmark and Share
  Featured Reviewer  
 
Data Mining (H.2.8 ... )
 
 
Social Networking (H.3.4 ... )
 
 
Social And Behavioral Sciences (J.4 )
 
Would you recommend this review?
yes
no
Other reviews under "Data Mining": Date
Feature selection and effective classifiers
Deogun J. (ed), Choubey S., Raghavan V. (ed), Sever H. (ed) Journal of the American Society for Information Science 49(5): 423-434, 1998. Type: Article
May 1 1999
Rule induction with extension matrices
Wu X. (ed) Journal of the American Society for Information Science 49(5): 435-454, 1998. Type: Article
Jul 1 1998
Predictive data mining
Weiss S., Indurkhya N., Morgan Kaufmann Publishers Inc., San Francisco, CA, 1998. Type: Book (9781558604032)
Feb 1 1999
more...

E-Mail This Printer-Friendly
Send Your Comments
Contact Us
Reproduction in whole or in part without permission is prohibited.   Copyright 1999-2024 ThinkLoud®
Terms of Use
| Privacy Policy