Uber: Where do people like to go?
When Uber arrived they changed the taxi-game forever. They provided a clean, comfortable way to hail a cab that brought a dying industry back to life. One of the cool things that Uber does is that it can provide very interesting data on traffic patterns and flow. You can observe where and when people like to travel. This information is helpful not only to Uber so they can position their drivers in the highest density areas, but also for cities as they try to manage the always annoying problem of traffic.
The study of how people move around is extremely important for city planning and can be modeled by developing a graph. The nodes are the starting and end points of a trip and the edge weight can be determined by how often that route is traveled. The city can then see the routes people are taking through the city and figure out how to manage their traffic lights, speed traps, and event planning.
Uber posted an article (https://newsroom.uber.com/2012/11/uberdata-mapping-a-citys-flow-using-ubers-ridership-data/) called “Mapping a City’s Flow Using #UberData”. In this article they show for nine U.S. cities (Boston, Chicago, DC, LA, New York, Philadelphia, San Diego, San Francisco, Seattle) what the Uber traffic flow looks like.
The Uber traffic flow for San Fransisco. (Screenshot from the google maps widget in the linked article).
This data is really interesting and shows how people tend to move through cities. It shows the probability of a ride starting in one area of the city and ending in another. Traffic patterns can be affected by a variety of variables like weather, events (Like a baseball game), and time of day. Since the graphics they provide don’t provide any of these variables from a city planners perspective they aren’t that useful. They become more of a “Oh that’s cool”.
Since this really interested me, I looked to see if anyone had done any more in depth analysis of Uber traffic and I found a blog post on MathWorks by Loren Shure (http://blogs.mathworks.com/loren/2014/09/06/analyzing-uber-ride-sharing-gps-data/) where he had access to the dataset (which is no longer available), and he did some more analysis. He was able to generate some really cool figures using MATLAB (Awesome, fun, and powerful language).
Uber Rides Depending on the Time.
He took the Uber ride data and compared it to the time that people were using it. You can tell from the second graph that a lot of people go out on Friday and Saturday night and take an Uber back in the wee hours of the morning.
Uber represents an interesting network to analyze. If you think of it as a large computer network with many different computers who are waiting around for tasks. As soon as a task shows up, one of those available computers takes it, runs the task, and then gets ready to take on another task. The Uber drivers would represent the computers in this network and the tasks would represent rides. Uber is basically generating a an extremely large network of drivers to handle all the potential rides of a city. This is really good. As long as Uber can keep it’s coverage of drivers high enough and in the right spots, there should always be a driver ready to take on a new ride. However, unlike a computer network, where the job can either get completed or not, an Uber ride is different. It has a human element where ride quality is very important. You may not care how your computer adds two numbers together, but you do care about things like your driver’s car, cleanliness, and path taken. So the path taken from point A to point B is very important to Uber.
Another interesting problem that Uber needs to solve is which driver is going to get paired with each ride. When a new ride comes in, Uber must determine which driver would be best suited to take on the ride and show that potential ride to the Uber drivers. Then it is up to the driver to take on the ride or not.
Uber presents many different, unique network problems and they are definitely doing a good job of solving them (Their still in business at least). The future brings in autonomous cars and Uber’s network even closer to the allegorical computer network mentioned above. And once that happens, they’ll be even more interesting network conundrums that Uber will have to solve.
Thanks to all my Uber drivers this summer for telling me about how it works behind the scenes and hopefully I sourced everything properly.
Sources:
https://newsroom.uber.com/2012/11/uberdata-mapping-a-citys-flow-using-ubers-ridership-data/