Capital Bikeshare’s open data is an invaluable tool that allows programmers and researchers to contribute to and learn from the system, making biking more accessible to residents and visitors in the Washington, D.C., area.
At November’s Mobility Lab-sponsored Transportation Techies meetup, CaBi Hack Night VI, programmers and Capital Bikeshare users presented tools they developed to improve the system, with perspectives ranging from personal convenience to planning efforts across an entire county.
Peeks at peak demand
Two presenters had developed models to predict Capital Bikeshare demand, which could ultimately help operators maximize the system’s efficiency.
Robert Kraig created a web app with a predictive model that seeks to use current data that helps planners make good decisions about how to develop the bikeshare system to mitigate demand. In designing it, Kraig explored variables that can inform which stations are most likely to be empty, and which docks are most likely to fill up throughout the day.
Kraig identified consistent trends among two groups: regular riders (Capital Bikeshare members and commuters) and casual riders (tourists). The two groups rode at predictable times, with commuters by the peak hours of the day, and casual riders by the day of the week. Using a predictive algorithm that calculates demand per hour in order to predict future usage, Kraig eventually created a web app that displays the likelihood of a bike being available at a station within the next 15 minutes.
Presenter Joe Paolicelli reached many of the same conclusions with his own model, including the use of the same random forest machine learning approach. He added hourly weather data and 10-day forecasts that, combined with historical data, helped his model reach about 95 percent accuracy and predict bikeshare usage to the hour 10 days into the future.
Manuel Darveau (above) of 8D, developer of the ubiquitous Spotcycle bikeshare mobile app, debuted a live demo of an upcoming Spotcycle app for the Apple Watch. With a few taps, wearers could check the docks closest to them for bikes. Spotcycle’s watch view also will include a ride timer and favorite-stations list to make navigating the system possible without pulling out a phone, and the app can even send bikeshare members an unlock code, helping those who do not have their key with them.
Journalist Andrew McGill created a similar application for the Pebble smartwatch, based on personal needs for his commuting experience: riding to his office’s station, only to find it full and forced to backtrack. Watch apps could help bikeshare users find the closest stations while on the go, making it more convenient to avoid being “dockblocked” before arriving.
Aysha Ruya Cohen’s graduate capstone project studied the factors surrounding bikeshare ridership in low-income and minority neighborhoods. In D.C., a disproportionate number of bikeshare members are white, college-educated, and make more than $50,000 per year. After analyzing dozens of variables for correlation with low ridership for coordination with bikeshare ridership, she found that D.C. could promote bikeshare to low-income communities better if it addressed streets that are dangerous for biking, membership costs, and perceived cultural differences.
Alex Rixey and Ranjani Prabhakar (at top), of Fehr & Peers, followed up on their presentation from Bike Hack Night V with the results of their Level of Traffic Stress study for Montgomery County, Maryland. Based on 2015 bikeshare origin and destination data, the team determined the shortest possible routes based on riders’ tolerance for stressful riding conditions.
They analyzed how riders with lower stress tolerance would have to follow significantly more indirect routes among many origin and destination pairs, creating an “indirectness factor” of how far out of the way that rider would have to go, as well as a threshold for added distance over which people will not ride to trade for comfort. There are even many cases where it is not possible to reach other stations without riding on more stressful roads. As a result, those areas with less stressful roads are more direct and connected, and therefore sport higher ridership levels. The study creates useful evidence to forecast potential ridership on a planning basis, providing Montgomery County with an idea for future stations and bicycling infrastructure that creates quality connections and increases bike ridership.
Jon Wergin, who has previously written about his project on Mobility Lab, presented his research of Capital Bikeshare users’ most popular routes based on GPS tracking information. While many models predict the shortest possible route between start and end points, using GPS trackers allows researchers to see where riders actually go, and better understand why they go those ways. The project was the first in the U.S. to employ special tracking devices attached to the bikeshare bikes themselves.
Using heat maps, Wergin visualized the most popular routes among the tracked rides. Casual riders stuck mostly to the sidewalks around the National Mall, and members largely followed streets with bike infrastructure, especially protected cycle tracks such as on 15th Street NW. Also telling, the most popular rides without bike lanes were on streets where bike lanes end abruptly, such as on 14th Street NW in Columbia Heights, indicating ideal areas to expand the bike network.
As Capital Bikeshare’s historical database continues to grow with its ridership base, planners and programmers will have even more quality information they can use to benefit riders and the region.
Photo: Top, Ranjani Prabhakar presents the results of the Montgomery County Level of Traffic Stress bikeshare study (M.V. Jantzen, Flickr). Middle, 8D calls in from Montreal to present their Spotcycle Apple Watch app (M.V. Jantzen, Flickr).
See more photos from CaBi Hack Night VI here.