Facebook Gephi Research

Post date: Mar 08, 2015 8:2:44 PM

Learners- Me, I guess!

The organization studied in my research was our elementary school Facebook page, which I manage. This is a page that I have felt needs a bit of help but never had/took the time to work on this and was not really sure what to do. This project seemed like a good time to find out some information about it in order to improve it. My question was how to get more people involved in it based on what is working so far and what is not working. I knew it was not getting quite the hits and connections that I wanted, but was not really sure of any details.

Data- Collected with NetVizz

I used a Facebook based program called NetVizz to get the data off the facebook page, which turned out to be no easy task until I figured out how to get the “code” for the page! I used the posts in this calendar year, 1/1/2015 to 3/7/2015 because any more than that ended up with too much data and took too long to get from Facebook. Then, I took out some things that were of no relevance to me for this study (such as gender) and it was ready to go. Each of the Facebook posts were a node and the likes comments, shares and connections from people were the edges. All the data was put into Gephi in order to make some sense of it and see if I could find a way to improve the use of our Facebook page.

Metrics- Done on Gephi

There are a variety of data visualizations and although they were interesting, and showed that there were occasions that pulled unusual people in, it wasn’t much help. There were definitely clusters where we had lots in connections but it was hard to tell what they were. Finally, after making many of these, I added the labels and enlarged it to see what they were. They were photos! All the things that were not in the clusters, were photos. That caused me to believe that although they get people to visit once, they are not all that productive in increasing our engagement with our school community. That surprised me so much that I went back and looked at them and at the responses to them. What I saw confirmed my suspicions. People will come see a photo, but those people often don't ever do anything but like or comment on that photo and we never see them again.

This is what it looks like when you first put it in Gephi. When you run what is called "Force Atlas" it causes the connected noted to be attracted to each other. It also moves the unconnected nodes apart. This creates clusters of connections. The second picture is in the middle of doing this and the next picture is the completed Force Atlas run with nodes moved into distinct clusters. This final one is much easier to "read" visually!

There is another Force Atlas called Force Atlas 2, so I ran it this time and added some color and got this visualization. Then I made it be all edges, with really small nodes and changed the color and got the green one below. I increased the size of the nodes and curved the edges and got the second one.

These were run with the ranking by degree (number of connections) so that the more connections they are, the darker the color.

Then I began looking at betweenness and centrality and changed the sizes to be according to their betweenness.

The labels made it too busy to read, but I was able to expand them and find out more information in order to see what was "popular" with people.

These are based on the modularity. I did it to get the cliques to show up more clearly.

Then there is the better visualization with the words and then the focused in picture of the words in the outliers. These turned out to all be photos! Here is the enlargement of the outlier area, and then I changed the tables to just show the actual type of node but I forgot to save that before I took a picture and there is NO going back with Gephi!!

These are In Degree and then In Degree that only start over 1, so you can see that there are many with only one that are no longer here. The next one is only above 2, which way removes a lot of it, making me think we need to work on getting more engagement!

The next set are the Out Degree- the second one is only those above 20! Very interesting, isn't it?!!

Intervention

If we want to get more people involved, we need to include more links that they find useful to students. These were the one with the most hits and the most engagement. The outliers are people looking at and probably liking photos, but not having anything else to do with the page. We are going to run some local paid ads to catch some families and we are going to ask teachers to invite anyone from our school with whom they are friends with on Facebook to like our page. I am going to ask teachers if they will repost and like a few things a week to increase traffic. I will also work to find links that have “perceived value” to the parents in order to get them to share them and link to them and increase our traffic.

I have to give credit to all of these sites for their information that helped me figure this out!!

https://dhs.stanford.edu/gephi-workshop/

http://gephi.github.io/users/

http://historyblogger.net/2013/08/17/getting-started-with-gephi/

http://www.martingrandjean.ch/introduction-to-network-visualization-gephi/

http://www.obviousleaks.org/step-by-step-gephi-tutorial/

https://persuasionradio.wordpress.com/2010/05/06/using-netvizz-gephi-to-analyze-a-facebook-network/

http://research.bowdoin.edu/digital-computational-studies/digital-computational-studies/excel-data-and-gephi-data-laboratory/

http://www.slideshare.net/gephi/gephi-toolkit-tutorialtoolkit

https://www.video2brain.com/en/lessons/the-facebook-graph-api

https://www.youtube.com/watch?v=XxH0Tm8NXik&feature=youtu.be

https://www.youtube.com/watch?v=i0W6XCSEfAE