Introduction
For lab three I choose to investigate flight pattern data. Typically, I use data that relates specifically to environmental issues or urban planning issues. This time, I choose flight pattern data mostly because it seemed interesting to map using network visualizations. My main question was how do airports relate to each other and what are a few of the major hubs? After playing around with this data set in Gephi, I have been able to address this question, and in addition, conjecture some interesting patterns relating to urban planning and the environment, though that was not my initial intention.
Through this module we learned about a broad variety of network visualization types and tips. For this lab I enjoyed spending time in Gephi, manipulating the data in different ways to get different display results. Overall, Gephi is a useful tool for creating network visualizations, though it is not as intuitive as some of the other programs we have used in lab 1 and 2.
Discussion
This was the first time I have every studied or created a network visualization. However, I love traveling and I have always been attracted to the flight maps that show the various flight patterns of different air travel trajectories such as the one below.
For this lab I wanted to try and mimic this type of visual. I’m not sure if the above flight map is technically a network but it has nodes and edges like the ones I learned about in this modules readings, and it inspired my lab topic. When starting any type of new project, it is important to get to know your data, so I first looked into the nodes and edges of this data set. The nodes represent airports and the edges represent flight paths. After loading this data in to Gephi I toggled between different layout options ran a few statistics and applied different filters to learn more about the flight path data I was using. Gephi made it easy to run statistics and apply filters, though it was sometimes challenging to adjust the labels to improve visibility. I knew the final result would be different than my inspiration (pictured above) since I am not mapping this network over an actual map. However, I tried to incorporate similar elements like curved edges and a similar color scheme in my final network.
Materials
The primary tool used in lab three is Gephi. The website for this tool can be found here: https://gephi.org The tool is designed to be open to the public however, it is not as user friendly as other freely downloadable software such as Tableau from Lab 2. This tool is set up to easily import excel doc and data in other common forms. The actual manipulation of the data requires more knowledge of how the tool works.
The data I used in this lab is the US Air97 data set which can be found here: https://github.com/gephi/gephi/wiki/Datasets. This shows the North American Transportation Atlas Data; the data set contains 2126 edges (flight paths) and 332 nodes (airports).
Methods/Process
Learning about Gephi:
To learn about Gephi I watched the demo video that Professor Sula posted on our canvas class page. In addition, spent several hours trying different functions of the program and created a variety of different versions of this network. The demo provided important step by step instructions for many elements of the program. Allowing myself ample time to experiment within the program was necessary for learning more about the program, the data set, and improving my network visualization.
Obtaining Data:
To create my network for lab 3 I looked through the three websites provided on canvas. I found a dataset that contains information on flight patterns and downloaded the data file from the website. It was already formatted in a way compatible with Gephi, making it easy to upload the data into the program.
Putting it all together:
After learning how to use Gephi and choosing a dataset to use, I loaded the dataset into Gephi and begin playing around with different layouts, filters, and statistics. One of my early attempts looked like this:
In this early attempt I had made the nodes and labels proportional by in-degree. I also created 3 modularity’s and assigned the edges to the in-degree node color. However, the labels were hard to read, the network was very dense, and the modules didn’t seem to tell a clear story.
Next, I tried adding a few columns to the data to sort the airports by geographic region or by international/regional status. However, this did not provide interesting statistics either. more Finally, I decided to use the expansion and contraction layouts in order to get the labels not to overlap, but my network was still so large that the labels where not legible:
Then I decided to apply a filter to the in-degree range. Originally the range spanned from 0-77. With the filter, this network only shows nodes with between 22-77 in-degrees:
This filter produced a much more legible network and a smaller scale of nodes and edges. I also tried changing the color schemes to be closer to a typical flight pattern map and got these results:
Result
The final network I created is below:
This Gephi network of flight patterns shows the relationship between different airports. The largest node is the Dallas/Fort Worth airport, indicating that of all the airports in this dataset, Dallas has the most flights flying in and out. The edges are colored based on in-degree, with magenta indicating more in-degrees (highest is 77 at O’hara) and white being least (8 at Seattle-Tacoma International and also General Edward Lawrence Logan).
Reflection
I definitely had a learning curve at the beginning, but the more I worked with Gephi the easier it was to trouble shoot issues and manipulate the data the way I intended. I enjoyed learning about networks but I am definitely more confident in making graphs and maps. I think Networks are a useful tool and I look forward to practicing with this tool in my future work.
References
Gephi download: https://gephi.org
US Air97 Data Set: https://github.com/gephi/gephi/wiki/Datasets