Network Analysis Basics (and applications to online networks)


Published on

Basics of network analysis, applications to the Web and social networks. Reference prepared for the Annenberg Program in Online Communities.

Download the presentation PDF from .

Published in: Education, Technology
No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide
  • Density=L/N(N-1)
  • More than half of the people using social networking sites, furthermore, receive and follow links to news items on a daily basis. Blogs are another medium in which news content is propagated. As it is an accepted norm for bloggers to link to their sources (Chin & Chignell, 2006; Ferdig & Trammell, 2004), information diffusion in the blogosphere can be tracked based on hyperlink patterns and time stamps.Taking into account those new trends, scholars have started studying the spread of topics through both social networking platforms (Oh, Susarla, & Tan, 2008) and blogs (Leskovec, Backstrom, & Kleinberg, 2009; Leskovec, McGlohon, Faloutsos, Glance, & Hurst, 2007). The two analytical approaches used to explore the online diffusion of media content involve threshold models (Valente, 1996) and cascade models (Cointet & Roth, 2009). In threshold models, an actor's decision to disseminate a topic is based on the proportion of their connections who have already started discussing the subject. In a cascade model, each time an actor is "infected" with a new topic there is a certain probability that the infection will spread to neighboring nodes.Epidemic models of diffusion are often used for online content as their robustness has been well established through long use in other scientific fields. Much of the research looking into the network flow of information is based on methods initially developed to model the spread of disease through interpersonal connections. Clinical research in epidemiology often uses the SIR (susceptible - infected - recovered) cycle to describe the stages through which a node may go. The same model has been adapted to study audience members and their exposure to media content (Leskovec, et al., 2007). In the online-diffusion version of SIR, users may become susceptible to a topic when it is suggested to them by a friend (either through a blog post or via a service like Twitter or Facebook). The person may then be infected with the topic, in that they write a post about it or publish it on a social networking platform. With this, the individual is considered to have recovered from the topic, although a relapse is possible when something new appears on the subject. Modeling the spread of media content through social networks has allowed researchers to understand topic life-cycles, spikes and declines (Cointet, Faure, & Roth, 2007; Gruhl, Guha, Liben-Nowell, & Tomkins, 2004; Leskovec, et al., 2007). It has also provided a way to explore patterns of influence and identify opinion leaders (Java, 2006; Nakajima, Tatemura, Hara, Tanaka, & Uemura, 2006).
  • Network Analysis Basics (and applications to online networks)

    1. 1. Network Analysis Basics and applications to online data<br />Katherine Ognyanova<br />University of Southern California<br />Prepared for the Annenberg Program <br />for Online Communities, 2010.<br />
    2. 2. Relational data <br />Node (actor, vertex, etc.) <br />Tie ( link, relation, edge, etc.) <br />
    3. 3. Internet<br />Internet Backbone: Each line is drawn between two nodes, representing two IP addresses. Source: Wikipedia<br />
    4. 4. Global Media Networks<br />Media corporations network. Source: Arsenault & Castells, 2008<br />
    5. 5. US Senate<br />US Senate cosponsorship network 1973-2004. Source: Fowler, 2006<br />
    6. 6. High-School Romance<br />Source: Easley, Kleinberg (2010) Networks, Crowds and Markets<br />
    7. 7. Online Social Networks<br />Twitter, 2010<br />
    8. 8. SNiF<br />Source:, 2010<br />
    9. 9. Basic Types of Networks<br /> HP<br /> MS<br /> Jill<br />John<br />Paul<br /> Jill<br />Paul<br />John<br />Kate<br />Kate<br />Jim<br />Jim<br /> Tom<br /> Jill<br /> Tom<br /> Jill<br />
    10. 10. A closer look at links<br />John<br /> Bob<br />John<br /> Bob<br />John<br /> Bob<br />John<br /> Bob<br />+<br />5<br />colleague<br />friend<br />
    11. 11. Distance in Networks<br /> B<br /> B<br /> B<br /> B<br /> A<br /> C<br /> E<br /> F<br /> A<br /> C<br /> E<br /> F<br /> A<br /> C<br /> E<br /> F<br /> A<br /> C<br /> E<br /> F<br /> D<br /> D<br /> D<br /> D<br />
    12. 12. A Closer Look at Nodes<br />
    13. 13. Node Types<br /> Isolate<br /> Liaison<br /> Gatekeeper<br /> Star<br /> Bridge<br />
    14. 14. Node Centrality<br />
    15. 15. Network Centralization & Density<br />Degree, closeness and betweenness are centrality measures for individual nodes.<br />Centralization is a network-level measure. It measures the degree to which an entire network is focused around a few central nodes. In a decentralized network, the links are more or less evenly distributed among nodes.<br />Centralization is calculated based on the differences in degree centrality between nodes divided by the maximum possible sum of differences.<br />Density : Ratio of the number of links to the number of possible links in the network<br />Size: The number of nodes in a network<br />
    16. 16. Example: Terrorist Networks<br />Source: Business 2.0 December 2001. Six Degrees of Mohamed Atta<br />
    17. 17. Social Network Measures & Mechanisms <br />
    18. 18. US Political Blogosphere<br />Linking patterns in the US political blogosphere. Source: Adamic & Glance (2005)<br />
    19. 19. What does network data look like?<br />Although there are other options (edge lists, node lists, etc.) network data is typically used in the form of a matrix .<br />Row X column Y is 1 if there is a link from X to Y - and 0 if there is no link.<br />The diagonal represents self-loops: links from X back to X. <br />C<br />D<br />A<br />B<br />
    20. 20. Using UCINET for Network Analysis<br />1<br />2<br /><ul><li>UCINET: download from this website. You get two months of free trial.
    21. 21. Data Input: - Direct: Click on (1) – copy and paste from Excel or enter manually a matrix. - Import: Data menu -> Import via Spreadsheet - > Full matrix w/ multiple sheets. For every network Ucinet will create two files: .##h and .##d. If you move them, move both.
    22. 22. Density : Network menu -> Cohesion -> Density
    23. 23. Reciprocity / Transitivity / Clustering : Network menu -> Cohesion -> Reciprocity / Transitivity / Clustering
    24. 24. Node centrality : Network -> Centrality -> Multiple measures (old)
    25. 25. Visualization: Press button (2) for NetDraw. </li></li></ul><li>Studying Social Structures Through Network Analysis<br />
    26. 26. Collecting Network Data<br />
    27. 27. Example: Snowball Sampling<br />Source: Valente (2010) Social Networks and Health<br />
    28. 28. The Web as a Network<br />
    29. 29. E-Mail and Discussion Forum Networks<br />Source: Welser, Gleave, Fisher & Smith, 2007<br />
    30. 30. Diffusion Through Online Social Networks<br />
    31. 31. Online Resources for Social Networks<br />How can we collect data about online networks?<br /><ul><li> Facebook:Name Gen Web by Bernie Hogan is a Facebook app that will collect data about your network and let you download it in a UCINET file format.
    32. 32. NodeXLis an add-on for network analysis in Excel. It can collect and analyze Twitter, Flickr and e-mail nets.
    33. 33. SNAP (Stanford Network Analysis Platform) has a data library with big network datasets from Google, Amazon, Wikipedia, social network sites, etc.
    34. 34. IssueCrawler, LexiURLand SocSciBotare some of the available solutions for crawling web pages and collecting hyperlink networks.</li>