Cloudcommuting

For the Cloud Commuting Midterm, Jiwon, Haylee and I analyzed CitiBike in order to identify communities of stations based on number of trips between stations. By analyzing bike trips between stations, would we discover a pattern of stations connected to each other due to a high volume of trips between them, and what would this pattern reveal about current usage of the CitiBike system?

To identify “CitiBikeHoods”, we wrote a Python program to calculate number of bike trips between stations from August 2014 (the most recent data available to us.) Thanks to Salem Al-Mansoori for helping us write this program (code at end of post!)

Once we had our data, we imported to Gephi, which is a software program which created a Network Graph of the stations and identified communities of stations. Initially, we were hoping to write a program using a Python module called NetworkX in order to create the network graph (allows for more control in generating graph and identifying communities from the data), but we had trouble writing the program so opted to use Gephi.

The tricky part of importing our data to Gephi was giving the data to Gephi in a format Gephi liked. We had to rename our Origin Station and Destination Station columns (our nodes), “Source” and “Target” and the column containing the number of bike trips between the stations we renamed “Edge” since it would be the weight of the line connecting our nodes (stations!)  We used a Network Graph to identify relationships between stations by grouping them into modules or “hoods.”

In the above graph, the outer yellow nodes are Brooklyn stations, the stations in the center are located in Lower Manhattan, the green stations are Midtown stations, the lower purple stations are Lower East Side/East Village while the upper purple stations are Murray Hill/FlatIron.

We then exported this data from Gephi in order to create a map using Mapbox to color in the stations according to the community the station belonged to. We chose to go with 63 communities but it might be better in the future to have less communities. View the code here. (Thanks to Adarsh Kosaru for helping us write code to color in our different communities with a randomly generated color in MapBox!)

What was immediately obvious in the mapping of communities was that Midtown is treated as one neighborhood, along with Brooklyn and Lower East Side, while other groupings of stations appeared to be random. In order to better understand the CitiBikeHoods, we compared to other maps of New York City. → Read more

For my first visualization of the CitiBike system data, I chose to focus on mapping out the path of an individual bike in the CitiBike system using data from May 2014. Jiwon, Billy and I met earlier in the week to review the available data and revise our midterm proposal for our analysis of CitiBike. We think it might be possible (although very difficult and not particularly scientific) that by following the path of a bike through the CitiBike system, we can figure out when a station has been refilled or is depleted of bikes. Each bike trip has a start and end station, and if the start station of the next trip doesn’t match the end station of the previous trip, then in theory, the bike was taken out of the system either for a) repair or b) to refill another station.

This is the path of one bike which visited a few stations before being removed from the system:

My visualization is pretty strange given I am drawing a line to connect the stations the bike started and ended at. I do not have data about the actual street route the bike took. It might be better to just draw a circle for each station the bike visited and a different color for the start and end stations rather than draw a line. However, this map does show that this bike was used a few times before being taken out of the system and moved elsewhere. → Read more