By Michael Davis
The video shows the evolution of the Hubway cycle hire network over time. Nodes represent the location of the cycle hire stations. Edges represent journeys between pairs of stations; the thicker the edge, the more journeys in that time period. Several things are evident from the visualisation:
- The network of hire stations was expanded in August 2012
- 3 or 4 months of data seems to be missing (Dec 2011-Mar 2012)
- In the early part of the dataset, there seem to be relatively few popular routes, but when the network was expanded, the number of popular routes blossomed.
The dynamic graph can be seen as a kind of hypothesis-generating tool, giving an insight into the data and allowing the user to pose questions about it. It is also useful for data cleaning, as it shows that the data is not all of equal quality.
Technical details: I created some software tools (using C++ and the Boost Graph Library) to turn the raw data into a dynamic graph in GEXF format. In the video, the dynamic graph is visualised using Gephi. (It is possible to make much slicker visualisations but I was focussed on getting the dynamic representation working; this is the best I could do before the deadline!)
The granularity of this visualisation is a week at a time over a year. The tools also allow the creation of finer-grained visualisations (e.g. showing how usage patterns vary over a single day).
Additional download: hubway_all.gephi