This network project used an interactive graphic showing an undirected social network of frequent associations between 62 dolphins in a community living off Doubful Sound, New Zealand. Dolphins live in pods and they develop various relationships with one another, so this is a good chance to see how they connect, as well as how many of them has a larger connection in this group than others.
A few network graph examples were being showed in class. One important lesson from those is not to make it too complicated. A network graph should focusing on providing information and allowing people to understand instead of making it fancy and colorful. It is crucial for this dolphin graph to be clear and precise.
The dataset was retrieved from CASOS Network Science Data. The original file is an XML file, with the nodes and edges data put together, so the data needs to be cleaned, then separated into two different files before imported to Gephi, the software for creating networks.
After deleting some extra columns and added a direction column for the edges dataset, the original data was exported to two different CSV files. They were then imported into Gephi, where the actual visualization step is going to take place.
The nodes file contains the ID number for each individual dolphins, the unique names that they have given by the scientists; the edges file has the source and target describing the relationship between dolphins, and this connection is undirected, with no weight being assigned.
Some calculations were done in order to generate some network related statistics so that Gephi can better visualize this network. Average degree can tell us on average how many connections each nodes have, network diameter shows how connected or disconnected the nodes are. In this case, it shows that any dolphin can be reached by 6 dolphins, or 6 jumps, which means that these dolphins are not very closely interconnected. In the end, graph density was being calculated, which is the connectivity of the graph. The dolphins has a density of 4.2%, which means that their relationship is not super intimate.
After the calculation comes the layout of the visualization. Force Atlas 2 was being used first to come up with a rough shape of the network. From this layout it’s clear to see that the dolphins don’t revolve around one single dolphins. The entire network is stretch pretty long, and even though there are a few dolphins have less connections, they build a bridge that helps connect their pod together. However, one problem of this layout is that some of nodes are very close and it’s hard to see the connecting lines between them. A second layout called expansion was being used, and since it expands the layout around its center, the nodes are more separated and a clearer connection was being shown.
After that, some style were being apply to the network. Degree were being used as an attribute to style the nodes, so that dolphins that are connected to more dolphins gets a bigger dot. A blue color scheme was added on top to the connectivity of the dolphins — the deeper the blue, the more connection the particular dolphins has. This style is also being applied to the edges lines.
In the preview section, borders were being removed from the nodes, labels showing each dolphins name were being displayed, with the same font size so that it is visible disregards the size of the node. A more simplified font was being applied.
The graph shows the basic structure of a dolphin pod. With these dolphins swim, eat, play together, and even protect each other, it is very interesting see that they are not as tightly connected as we might think. The network shows us that in this pod of 62 dolphins, only around 5 dolphins are connected to 10 or more others. Most of them only interact with less than 3 dolphins. This one big pod of dolphins are also 2 smaller communities in one, with around 5 dolphins being the bridge between them. With a low interconnection, it is amazing to think about how they collaborate with each other.
Of course, many information is not included in this graph. For example, since the dataset is undirected, we don’t know if their connections are one way or both way. We also don’t know if the connection is positive or negative, since if one dolphin dislike another one, we could still say they have a connection. We also don’t know if dolphins who have more connections means they are more popular, or if they are older, or more experienced. These information can make this network more interesting, and provide us with more understanding of how a dolphin pod function.