Mobility and Bike sharing
The quality of life in a city is measured by several parameters, one of them being the time to move around. This is one of the biggest issues in any major city including NYC. The city network of bus, trains and the large availability of cabs, features areas of excellent as well as very poor connectivity, where life without a car can be critical. CitiBike is the recently launched bike-sharing program of New York City with the hope of making cycling an alternative and sustainable form of transportation as already happens in other major cities in the world such as Paris, Montreal and DC.
The system has enjoyed immense popularity but has been plagued by several problems such as in first place the lack of resources and poor maintenance.
The aim of this project is to make use of available data from CitiBike to optimize and guide the expansion in new areas, which have been already planned. The system is expected to grow by 20% by the end of 2015  with 90 new stations coming to several neighborhoods in Brooklyn, Queens and Uptown Manhattan.
My project has long-term goal of building a predictive model of the distribution of bike stations in those new areas that is optimal in the sense that it covers the largest area but takes into consideration the heterogeneity due to the presence of landmarks or hub of connectivity such as train or bus stations.
And also understand what are the most “active” boundaries meaning the most profitable direction of expansion. To begin with, this requires understanding the activity of each stations based on its neighborhood, what is the typical trip of the type of user (tourist or commuter). Moreover, a natural question is if the bike network is competing or integrating the public transportation network MTA. This is crucial to plan the next generation of Citibike.
As a first step I have analyzed the datasets for trip histories including information such as trip duration, starting and ending location based on user type over a period of 18 months between July 2013 and December 2014. This results show that the geometry of the city is crucial and can be recovered from the average trip properties.
I present preliminary results divided in two sets of plots:
1) Real-space imaging of station activity, averaged over daytime and for different kind of users shows heterogeneity and asymmetry of bike-usage.
This result shows that Manhattan is currently the main center of activity where most of the trips are concentrated. There are two types of users: subscribers with an annual contract and customers with a daily or weekly pass, and they use the service in a very different way. The former, being mostly commuters, are moving within the main transportation hubs in midtown. Their activity is one order of magnitude higher than that of the Customers. In particular they move around midtown with high peak of activity in the stations around Penn Station, Grand Central and Port Authority. This means that the distribution of stations in those areas has to be higher than in residential area in order to be optimal. Those hubs are also asymmetrically most frequently used as the starting point of a trip (colored in blue on the map). This requires supplying a continuous flow of bikes on the opposite direction. The customers are instead more evenly exploring Manhattan and Brooklyn too with a big activity on touristic sights Central Park at 59th Street or World Trade Center as accumulation points. Thus, customers tend to gather in the same places using the same ending stations, showed in red in the map. This is showing that users have different needs depending on their type and this influences the stations activity. Moreover, it gives us a lot of information to plan the next locations. In particular the possibility of expansion around Central Park seems more and more a very profitable direction with lots of tourists visiting the area .
2) Trip statistics, including probability distribution of trip length and trip direction, gives insights in the geometry of the space.
This set of result is very interesting since it shows the emerging of geographical boundaries of Manhattan from the bike activity. The average length of a trip is 2.02 Km for Customer and 1.74405 for Subscribers with a standard deviation that is comparable. This is essentially set by the west-to-east extension of Manhattan and the distribution of angles reveals something on the geometry itself. The most probable direction of the trips is approximately the angle that the Manhattan street grid is rotated from the north-south axis, 29 degrees. In other words this reveals that most of the bike activity is along the direction of the streets that is not served by the subway. This is interesting since this is traditionally the most difficult shift to make in NYC suggesting that the bike system is not in competition with the train but rather integrate it with it. This may suggest useful ways to plan the distribution of stations both in Manhattan and in Brooklyn.