Final Project

• The final projects you can do them alone, or in teams of 2 people.
• If you do not have any partner and you prefer to do the project with someone else, mail to Aris to try to match you. The earlier you mail, the higher the chances that will be to find a partner.
• You have two project options. One is a literature survey on a topic and the second is a programming project.

• You can start with a paper from the proceedings of
  • KDD 2009
  • KDD 2008
  • WWW 2010
  • WWW 2009
  • WWW 2008
  • WSDM 2010
  • WSDM 2009
  that is relevant to the class material.
• You should read some more papers related to the topic of the paper. For that you should check the previous work (found by the citations at the end of the paper), and any new work (you can use Google scholar for that). We expect that you will read at least three more papers to include in the survey.
• You should prepare the survey, which should be a document of about 20 pages if you use the solution format, that describes the area related to the paper and a list of related papers that you have found.
• At the exam you should hand in the survey and do a presentation of about 20 minutes.
• Before starting working on the paper, you should send an email to Aris and Ilaria describing the initial paper that you found and the area for which you will write the survey.
• If you have any questions about anything related to the survey, mail to Aris or Ilaria.

• You should choose some programming problem related to the class material, such as mining of social networks, analysis of query logs, computational advertising, etc.
• When you find a topic send an email to Aris and Ilaria describing your plan, the size of data that you expect to work on (e.g., how many nodes do you expect your social network or the click graph to have?) to make sure it is doable and that suffices for the coursse.
• At the exam you should hand in the code and do a presentation of about 20 minutes.
• Some concrete ideas for project topic follow at the end, but feel free to propose something else.
• If you have any questions about anything related to the project, mail to Aris or Ilaria.

1. Analysis of foruming.it.
   • Crawl foruming.it.
   • Create the social network of users who participate. Find a reasonable definition of when two users are connected.
   • Compute statistics of the network such as degree distribution, clustering coefficients, diameters, etc.
2. Detect communities on some social network (e.g, IMDB), using the divisive algorithm of Newman and Girvan that we did in class. Find the best level using the modularity measure.
3. Crawl a part of the Facebook network. Gather friends, create the social network of people who participate.Compute statistics such as degree distribution, clustering coefficient, diameter, etc.
4. Implement an algorithm for community detection based on the modularity measure. Check with Aris for ideas about a specific algorithm.
5. Implement the simple model for burst detection (the version with the two states low and high) from http://www.cs.cornell.edu/home/kleinber/bhs.pdf.
6. Perform query recommendation based on random walks on the click-graph as described in http://research.microsoft.com/en-us/um/people/nickcr/pubs/craswell_sigir07.pdf
7. Create a facebook application that performs something interesting. Especially welcome are applications that give incentives to become viral.