Explore the behaviours of New Yorkers through Foursquare check-ins
Woojin Kim, Arushi Arora, Azam Alfayez · December 10, 2015
New York City is arguably the best-known city in the world. The most populated city in the United States, it is home to over 8 million citizens as well as hosting over 56 million tourists just in 2014. As such, New York is one of the most popular cities on Foursquare, where users can "check in" and share their locations with their phone to share with friends, earn badges, and other rewards in the website. The Foursquare dataset made available by Dingqi Yang contains location, time, and venue information for over 2 million check-ins in New York recorded over nearly a year from April 2012 to February 2013. In this project, we attempt to explore the behaviours of Foursquare users in New York.
One day in New York. Every check-in condensed and visualized as one day of check-ins
How do check-ins vary throughout the day? How much can we learn about our days from our check-in patterns? The two million check-ins in the dataset resulted in roughly over 180,000 check-ins events each month, allowing us to have a granular look at users' behaviours through many different timeframes of the year. Try different combinations of months and days of the week below to see the number of check-ins at every hour of the day.
As expected, there are three clear peaks showing coffee/breakfast, lunch, and dinner times in almost every timeframe, then a tapering off of late night bar check-ins. There isn't a particularly discernable pattern through the different months, but the difference between the days of the week is quite defined.
On the daily histogram above, the most striking differences were seen between the days of the week, particularly between the weekdays and the weekend. Here, we see all the days of the week at the same time. Hover over the lines to explore highlight a specific day.
The three peaks for coffee, lunch, and dinner were clearly defined for the weekdays. For the weekends, however, there isn't a clear peak throughout the day; instead, it is a more gradual rise of check-ins, peaking sometime after noon. The wake-up times are noticeably later for the weekends than the weekdays as well. Interestingly, while the wake-up behaviour looks identical for both of the weekend days, the Saturday night pattern is more like that of Friday.
Not surprisingly, people seem the most deperate for coffee on Monday mornings. Perhaps in preparation for Friday, Thursdays have the lowest peaks among all the weekdays.
The category information included in the dataset also allows us to see the differences in check-in patterns between different venue categories. The following series of graphs show the number of check-ins throughout the day for six different kinds of venues.
By far the most popular category for check-ins in New York City is bars. As expected, the number of check-ins explodes as typical working hours end at 5 pm, peaking at 11 pm. The last call in New York is at 4 am, accordingly, there is very sharp decling leading up to it. There's still a high number of baseline check-ins throughout the day. As many bars double as restaurants and coffee shops during the day, it's likely that many daytime check-ins are as a result of this.
Coffee shops and subways both show a high morning peak, but with an earlier peak for the subway, presumably as people get coffee after getting to their workplaces. Another category of interest is gyms and fitness centres, showing two distinct peaks for some people trying to fit gym before work and others after. It's interesting to see that while subway check-ins are great for visualizing when people get to work, other categories like bars and gyms are more useful at visualizing when people leave work.
This chart depicts the top 12 check-in categories for each borough as percentage of total check-ins for that borough. As we saw above, it is no surprise that bars are the most popular check-in category in Manhattan followed by the office. For all the other boroughs, homes constitute the highest proportion of check-ins followed by subway check-ins, suggesting that generally Manhattan is the place for work, while other boroughs are more popular for residences. Here we can also see borough specific activities like airport check-ins for Queens thanks to JFK & LaGuardia and ferry check-ins for Staten Island thanks to the Staten Island Ferry.
As our social networking use continues to grow, it's also incredible to see the amount of information collected about our daily lives. While tools like Foursquare may have been for sharing check-ins in the beginning, as we see above, its data provide a fascinating insight into people's behaviours, far beyond what surveys only decades ago may have provided.
This project was created by Woojin Kim, Arushi Arora, and Azam Alfayez for the Storytelling with Data course at Columbia University.
We would like to thank Jonathan Soma for the amazing tutorials and, most of all, a fantastic course. This project makes use of D3, CartoDB, Bootstrap, and jQuery. Foursquare check-in dataset courtesy of Dingqi Yang (PDF).