Network Analysis

In this week's lectures we learned the basics of Network Analysis. This is a very fascinating area of study and although I have some exposure to network analysis from previous course work I'm just thrilled by the possibilities and am eager to dive in. As internet usage proliferates further and further and proliferates, as we advance toward a holistic internet of things, we are increasingly engaging with others in new and meaningful ways and the networks we form can yield tremendous insight into our lives. To be able to tap those networks and analysis the data therein has huge implications not just for business but all analytical fields.

In our first lecture we discussed the basics of what networks are. Defined as a collection of entities and the relationships among them, networks quite literally are our every interaction quantified into graph form. Each individual in a network, be they a person or a website, is a vertex, and every interaction or relationship is an edge. These vertices and edges come together to form the network which can be mapped.

An example of a Network map, using Star Wars : D! 

If we look at the above image on the far right, we can see Kylo Ren and Supreme Leader Snoke; they're both vertices and the relationship between them is the edge. Beyond the simple example above networks can also be directed or undirected and weighted or unweighted. Directed networks indicate directionality in a relationship (edge). The above network is undirected as it does not show in which direction, if any, the relationship flows. Weighted networks show if any individuals (vertex) are more important than others. The above network is weighted, as it shows the more important links as larger circles. There are also single node and two node networks; single node, like the network above, one type of individual vertex and the two node network has two different types.

In the second lecture we discussed the different ways of visualizing networks, of which there are many. In the third lecture, we discussed the properties of networks in regards to analysis. There are four kinds of ways to measure what is called centrality. Centrality is essentially how important an individual node (vertex) is in the network. We can measure centrality by degree, betweenness, closeness, and eigenvector. Degree is how many nodes an individual node can reach directly. Betweenness is the measure of the shortest route between two other nodes through a specific node. Closeness is how fast everyone in the network can be reached from a specific node. Finally, eigenvector tells how closely a specific node is connected to other nodes that are well-connected. Using these four measures we can crack the code of how the network is formed and interacts.

In my opening paragraph I mentioned how network analysis has broad applications and even a quick google search will show that to be true. Network analysis possibilities start with the mundane; there are tools available to visual one's personal LinkedIn or Facebook. They grow from there. Network analysis can be applied to ancient works, as this article about using network analysis to analyze the major characters in the Irish myth cycle shows. Network analysis can be used to visualize and analyze the usage of a hashtag on twitter, or even help predict the use of weapons of mass destruction. The applications are incredibly vast, and this introduction to the world of network analytics has been very enlightening and exciting.