Network visualizations are used to visually demonstrate relationships between entities that aren’t neatly structured. Networks are useful in helping to identify connections and links that form significant relationships and study the significance of these relationships. A network can be used to study many things, such as social media connections, connections of scholars, as well as things such as web traffic. Being able to actually look at these connections in a network visualization can allow a researcher to analyze the relationships in a variety of ways.

For our class networking lab I chose to work with a data set from the Carnegie Mellon University Center for Computational Analysis of Social and Organizational Systems (CASOS). The data consists of the 136 multiple directors of the 108 largest joint stock companies in Scotland for the years 1904-1905. During this time, joint stock companies were responsible for running the major industrial corporate giants. Many members of the board of directors of these companies served on multiple boards, creating links between them. By visualizing this data in a network, I believed it would be possible to see which companies had the most connections.

The data was presented in XML format, so I first had to convert it to a CSV format before I could load it into Gephi. Once the data was correctly set up as an edge table, I imported it into Gephi and began to analyze the network. The result is a directed network with 142 nodes and 648 edges. The average degree is 4.563, the average weighted degree is 5.042, the graph density is 0.032, network diameter is 8, and modularity is 0.443.

The first realization I had was that the data set split up the full names of the joint stock companies into separate words rather than as full entities. As a result the node with the highest degree of 82 is “&”. The five nodes with the highest degrees are: &, railway, British, of, and Scotland. The five nodes with the lowest degrees are: bridge, burmah, subway, sulpher, and devan.

While after visualizing the network I realized that it would not enable me to properly analyze the connections between joint stock companies, it could be used for analyzing wording in naming of joint stock companies. While I was not able to analyze what I initially set out to, the data still formed an interesting network.

To group the network, I ran a modularity with a resolution of 1, but this produced too many groups. When represented in the network, it was very difficult to differentiate between them. I ran a modularity a second time with a resolution of 1.5 and this created 6 groups which was much more manageable. I used this modularity to partition my nodes and the resulting network is much more visually comprehensible than the original modularity.

While it is difficult to draw any significant conclusions from this network, I did manage to make some noteworthy observations. First, the network does not include any isolates, each node has at least 2 edges. However, I believe this is due to the way the data was structured. As the joint stock company names were split up by word, each word would at least be connected to the other words in that company name. Second, the term “British” has a higher degree than either “Scotland” or “Scottish”. This may not be significant, but could prove interesting for history of the organization of industry in the United Kingdom. Finally, despite “&” having a higher degree than “railway,” the “railway” node is located centrally in the network while “&” is far from the center.
While I was not able to analyze the network in the way that I originally intended when I found the data set, I was still able to gain hands-on experience working with network visualization. Networks can be very intricate, and using Gephi was very difficult, particularly at first. I would like to use the software again in the future but hopefully with a data set that would allow me to complete a more meaningful analysis. After completing this lab, I feel confident that I would be able to use Gephi for future analyses.