In my work and play I have analyzed many networks using R code, as well as a myriad of third-party tools to answer a number of technical questions about data. Some of this data is in the form of relationships, some are in the form of social media shares and retweets, and some are based on publicly available structural meta data. Below you will find a few examples of my findings and visualizations.
Daesh – ISIS – ISIL: A Terror Group’s Social Media Network
In 2015, I conducted an analysis for NATO STRATCOM Center of Excellence on how Daesh disseminates propaganda online and how they make use of social media. The network showed here is of a sample drawn from a backcloth network (friend/follow) – the original network was made of roughly 160,000 nodes and roughly 500,000 edges, so it was very difficult to export to a web. Therefore, a sample of 3000 nodes and their edges were provided. the conclusions of the analysis itself are numerous and can be read by downloading the report from the knowledge store.
Political Analysis – The Obama-Ryan Debate
In 2012, the US presidential elections captured the essence of what big data and social media can do for business and political campaigns. They also provided incredible insight into how grass roots organizations manifest themselves on social media.
This particular network study was conducted on twitter, hashtag #Healthcare in October of 2012 – only 1 month before the general elections. I was lucky to have been able to collect all the data and not just a sample of it (as you know, since then the twitter api has changed allowing only a sample of the data to be collected for free.)
Earlier in the day, Paul Ryan (his twitter account), the vice presidential nominee for the republican party, tweeted a message using the hashtag healthcare. This was immediately retweeted by his direct followers, and prompted a response by Barack Obama (twitter account).
There are 3 interesting findings from this map and analysis:
The first is that because of Ryan’s only recent entry into the political campaign he was unable to get his message to spread much beyond his first degree followers. It could also mean that the RNC did not have an active campaign to integrate social media strategy that included Ryan’s twitter account. The second is that Obama’s social media strategy was much more developed because of his (campaign’s) ability to ensure that his messages were being actively retweeted to second, third, and even fourth degree followers, garnering much more attention and allowing for multiple imprints on the same followers.
Finally, and perhaps the most important learning, is that without utilizing sentiment/semantic analysis one can easily identify political leanings and affiliations through structural analysis alone, which makes this map some very tasty food for thought.
Choices & Demographics – The State Department’s Fulbright Program
The United States Fulbright program is an initiative designed to promote scholarly collaboration between American scientists and the rest of the world. Every year, US scholars of all ages, disciplines, and academic proficiency apply to this program, which is headed by the State Department in order to experience a sponsored/semi-sponsored international academic experience. Promoting positive relations, exchanging ideas, and expanding the mind of the scholar are only some of the goals that this program aims to achieve.
As expected of any highly competitive program, often times there are only a limited number of spots available for any given field, country, institution or some combination thereof. Therefore, the State Department’s administrative liaison, the IIE, would make suggestions for other countries within the same region of the applicant’s first choice. For example, if an applicant chooses Germany as their first choice, but all the spots available for Germany have been filled, then IIE would recommend France, or the UK.
My analysis helped to change this thinking.
Running a clustering analysis on the choices made by applicants can be classified as 2-mode network analysis, but it also represents 3 data modes: the number of choices made by applicants, academic institutions to which they belong, and country of choice.
The result was unexpected. Applicants that were tied together by similar clusters did not choose a country in the same region when the first region choice was rejected. In fact, we found clusters of countries to which applicants from certain institutions highly favored. This and other follow-up analysis allowed us to conclude that suggesting a country from the same region may not be the optimal solution.
Analyzing Gamer Choices – EVE Online, the World’s Most Popular MMO
Like everyone, sometimes I love combining several of my passions into one activity. I do it often when I snack and watch a movie, when I exercise and listen to music, and in this case when I analyze a massively multi-player online sandbox that I enjoy as a gamer.
This network analysis represents the structural design of space systems within the even universe. It lets any gamer not only see a clearer vision of the connections between systems, but also allows more efficient game decisions to be made.
The network map makes it easy for a fellow pilot to identify trade, battle, and transportation opportunities previously unseen without a solid graphical representation of the structural attributes of the game.