What Can Open Data Tell Us About New York Taxi Journeys?

Mobility Data Stories Geodata

August 12, 2015

Reading time: 3 min

Read this blog post and explore data about yellow taxi journeys in New York City.

Last week, New York City opened data from millions of taxi trips. 165,114,361 to be more precise. At a time when war rages between taxis and Uber, just a few days after the clash between NYC's mayor DeBlasio and Uber, it looked like a good a idea to play with this data by uploading them into the Opendatasoft platform and displaying them on a map and as well as on graphs and charts!

 

South of Manhattan, NYC, Heatmap of Taxi trip pickups and Subway Entrances locations
South of Manhattan, NYC, Heatmap of Taxi trip pickups and Subway Entrances location

 

Some Facts

As we could expect, there are more and more taxi trips during the week, peeking on Saturday.

 

Number of trips, sum of distance and total fare by weekday
Number of trips, sum of distance and total fare by day of the week

 

The month by month evolution seems to show more taxi trips during Spring and Fall:

 

Number of trips, sum of distance and total fare by month
Number of trips, sum of distance and total fare by month

 

It would be interesting to wait for 2015 data to see if there is a real pattern though.

Reverse Engineering of Taxi Pricing

Here are, given a trip's fare amount, the average distance and average duration of the trip:

 

Average distance and average duration given a fare amount
Average distance and average duration given a fare amount

 

It may be worth trying to recreate a pricing model from this data and create an app that tells you when you are way too far the average, or by filtering on the pickup location checking if part of the New York City population is being disadvantaged as some people claim.

About Data Quality

For a dataset this huge, there are not a lot of errors or bad data, but absolutely clean and perfect datasets are really unusual. Data visualization and mapping are good ways to find some incorrect data, especially when you have 160M rows to check! The first, and most basic, example is bad geographical coordinates:

 

North America scope - January 2014
North America scope - January 2014

 

I know Uber has launched Uber Boat in Istanbul but most of the markers far from NYC are probably incorrectly located. With the same idea, we can see that there are both very long trips - 13 days - that may need more investigations, and trips with negative duration, that don't need any investigation:

 

Number of trips by trip's duration
Number of trips by trip's duration 

Smart Cities 101

The harder the funnier, we've created a heatmap of every pickup locations in 2014 (~160 million, remember) and add the Subway entrances as a layer. It's pretty amazing to navigate that easily in those huge datasets!

By implementing some route calculation between the subway stations, we could compare every taxi trip with a public transport trip and understand better how people behave, why, and what the city can do to improve their lives. That would be a nice first step in the development of a Smart City:

 

taxi_opendatasoft_demo

Read more from this author

Topics

Mobility Data Stories Geodata

Subscribe to
Our Newsletters