Determining mutual followers and cluster coefficients of a user's Twitter friends

Feb 11, 2015 at 6:13 AM
Edited Feb 11, 2015 at 6:14 AM
Hey all,

I'm not technical, but a do a lot of data analysis, and use Gephi for analyzing community data on users of our own product.

I need to do 3 things, and I can't find a reasonable solution with NodeXL, though I believe it's totally possible:
  1. I need to examine all of the followers of users that I follow (there are <200 users I follow), to look for mutual followers among them.
  2. I need to get a cluster coefficient of users that follow the people I follow on Twitter.
  3. I want to be able to put this into Gephi and run Force Atlas to examine all of the followers of the people I follow, and how they are connected. I'm completely comfortable in Gephi.
Can someone point me in the right direction? Thank you in advance!
Coordinator
Feb 11, 2015 at 2:46 PM
Edited Feb 11, 2015 at 2:47 PM
Hello!

Thank you for the interest in NodeXL!
  1. I need to examine all of the followers of users that I follow (there are <200 users I follow), to look for mutual followers among them.

    Access to "follower" edges in Twitter is strictly limited. Please have a look at:
      http://www.connectedaction.net/2013/06/11/over-the-edge-twitter-api-1-1-makes-follows-edges-hard-to-get/
    
    However, NodeXL can collect the follower relationships (over time) via the NodeXL>Data>Import>From Twitter Users network....
  2. I need to get a cluster coefficient of users that follow the people I follow on Twitter.

    Clustering coefficient is a vertex level metric that measures the density of connections among an "ego's" 1.5 degree local network.

    Density is a network level metric that measures the percentage of observed connections out of the total population of all possible connections in a whole network.

    It sounds like you want the density of your local follower network, not the clustering coefficient. This metric is reported in the "Overall Metrics" worksheet in NodeXL.
  3. I want to be able to put this into Gephi and run Force Atlas to examine all of the followers of the people I follow, and how they are connected. I'm completely comfortable in Gephi.

    You can export NodeXL data sets to GraphML and import GraphML into Gephi. See NodeXL>Data>Export>To GraphML file...
Regards,

Marc
Feb 11, 2015 at 3:30 PM
Thanks Marc,

I think Ive just gotten used to using certain terms improperly ;)

You may be familiar with Twitter's Magic Recs, an experiment they run to personalize notifications and DMs for users. One thibg Magic Recs started doing is notify you if 2 or more people you follow favorite, retweet or follow the same person. In a roundabout way, I'm trying to look at a cluster of Twitter users to understand roughly how many of these notifications will be sent if the entire cluster performs the same action on Twitter.

1. This requires me to examine mutual followers among the cluster (how many followers of everyone in the cluster follows at least 2 people in the cluster). How do I get this? NodeXL seems to focus on interactions (mentions, retweets) on Twitter, and the relationships in those actions, rather than purely the following/ follower relationships themselves. I don't care about interactions, only relationships. Let me know if I'm making sense.

2. Density will is also important, and you've explained that well :)

I was very excited to see you reply to my last email! It's a huge help to me. Looking forward to hearing from you.

Take care,
Ryan

Sent from Outlook




On Wed, Feb 11, 2015 at 7:47 AM -0800, "MarcSmith" <[email removed]> wrote:

From: MarcSmith

Hello!

Thank you for the interest in NodeXL!
  1. I need to examine all of the followers of users that I follow (there are <200 users I follow), to look for mutual followers among them.

    Access to "follower" edges in Twitter is strictly limited. Please have a look at:
      http://www.connectedaction.net/2013/06/11/over-the-edge-twitter-api-1-1-makes-follows-edges-hard-to-get/
    
    However, NodeXL can collect the follower relationships (over time) via the NodeXL>Data>Import>From Twitter Users network....
  2. I need to get a cluster coefficient of users that follow the people I follow on Twitter.

    Clustering coefficient is a vertex level metric that measures the density of connections among an "ego's" 1.5 degree local network.

    Density is a network level metric that measures the percentage of observed connections out of the total population of all possible connections in a whole network.

    It sounds like you want the density of your local follower network, not the clustering coefficient. This metric is reported in the "Overall Metrics" worksheet in NodeXL.
  3. I want to be able to put this into Gephi and run Force Atlas to examine all of the followers of the people I follow, and how they are connected. I'm completely comfortable in Gephi.

    You can export NodeXL data sets to GraphML and import GraphML into Gephi. See NodeXL>Data>Export>To GraphML file...
Regards,

Marc