I am interested in researching the predictive power of combining social network analytics (eigenvector & betweenness centralities, community detection, collaborative filtering, etc.) with traditional data mining techniques (random forest, K-NN, principal component analysis, etc.)

I have run into problems finding open source data sets that contain both relationship data for the SNA as well as attribute data for the traditional algorithms, and finally some attribute of interest worth modelling (accepted offer, $ spent, etc.)

Could anyone offer suggestions?

  • i'm really clueless here. got a link that recaps all that? lots of blogs implement xfn, so there's your social data...but i'm having a hard time tying that to attribute data...because i do not know what i am talking about.
    – albert
    May 10 '14 at 15:51

I'm not sure if you care what domain the data are from, but I suspect the Add Health dataset may have what you're looking for. It contains a bunch of variables on health and child wellbeing and includes data on respondents' school-based networks.

  • I was hoping for something without the privacy terms that the Add Health set includes, but it is better than anything I've seen yet, so I accept your answer. Thanks. May 17 '14 at 17:34

You could check out our data sets. This is synthetic data generated to match the census data as closely as possible, without having any privacy concerns. It contains both demographics, household relationships, and social contact networks.

  • 1
    I would up-vote but I blew all my reputation on that bounty :). Jun 11 '14 at 22:12

Inploid is a social question & answer website. Users can follow others. Each user is associated with a reputability score which is determined by the feedback of others about the questions and answers of the user. Each user can also specify interest in topics. The data is crawled in June 2017 and consist of 39,750 nodes and 57,276 directed links between them.

For each user, reputability scores and top five topics are included in the dataset. The dataset might be little sparse due to registered but inactive users, so I recommend removing inactive users (based on degree, empty attributes, or by finding largest connecting components). The dataset is available here.

It's been a long time since you posted the question but hope this dataset becomes useful for the ones who might be looking for an attributed social network dataset.


Another option would be using anonymized ego network datasets from the Stanford Large Network Dataset Collection. The data contain a graph of undirected relationships and features observed for each vertex (e.g. profile information). The samples are from large social media platforms:

  • Facebook (4039 vertices, 88234 edges)
  • Google+ (107614 vertices, 13673453 edges)
  • Twitter (81306 vertices, 1768149 edges)

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.