A report containing information about our process in this project can be found here.
The downloaded data from the TLC contained over a million observations for each service dataset, so we decided to limit the amount of data for our analysis. First, we filtered our data to only contain taxis or hired vehicles going into Manhattan. Our logic behind this was that this would reflect the general habits of New Yorkers, as most of the main attractions of New York City reside in the borough of Manhattan.
We also decided to sample the data to limit the number of observations for each type of taxi service. We used a ten percent sample for both yellow and for-hire taxis, and a twenty percent sample for green taxis since there was less data available. We merged all of these samples into one dataset that would represent data for the entirety of Valentine’s Day.
Upon an exploratory analysis of this dataset, we found that there were few rides coming from Staten Island into Manhattan. Therefore, we decided to focus on the other four boroughs for analysis, since there was not enough data from Staten Island to come up with meaningful interpretations.
With this cleaning, we made graphical representations, such as heatmaps, time distribution graphs and inferences on congestion to help illustrate where people are going and how far they are willing to travel for a night out on Valentine’s Day.
The time distribution graphs showed that 6-9 PM were some of the most popular hours in all of the boroughs on Valentine’s Day. 8 AM was the busiest hour in Brooklyn and the Bronx, but after that, dinner hours were busiest in those two boroughs. The stacked histograms also showed that yellow cabs accounted for nearly half the rides in Manhattan, but were not nearly as popular in the other boroughs. There were still a decent amount of pickups with yellow cabs in Queens and Brooklyn, but in the Bronx, for-hire vehicles constituted the majority of pickups.
Looking at the time distribution graphs for just the dinner hours, 7-8 PM seemed to be the most popular time to get picked up for dinner, but the difference between the three hours was minimal for pickups. In general, on Valentine’s Day, people were using cab services to get to work in the morning or to travel around dinner time.
Based off the heatmaps, each borough had their neighborhoods and zones of preference. Within Manhattan, many of the rides stayed within the same zone, implying that many people did not travel very far in taxi or in an Uber or Lyft on Valentine’s Day. Neighborhoods that tended to stay within their starting zone include the Upper East Side zones, the Harlem zones, and the Midtown zones.
For Brooklyn and Bronx, proximity to certain neighborhoods played a major factor. The more frequent rides from Brooklyn originated from Williamsburg and DUMBO/Vinegar Hill, which is geographically closer to Manhattan than many other Brooklyn neighborhoods. In addition, these pick-ups often were often dropped off in closer Manhattan neighborhoods such as the East Village.
The frequent rides from the Bronx also reflected this, as most of the Manhattan destinations were in Northern Manhattan zones such as Inwood, the Harlem zones, and Washington Heights North and South.
Housing both LaGuardia and JFK Airports and being most west of Manhattan, Queens did not have many rides going into Manhattan. Majority of the rides going into Manhattan from Queens originated from the airports. Most of the rides from the airports ended up in Times Square/Theater District or the Midtown zones.
Manhattan to Manhattan travel is pretty consistent with its congestion, whereas it fluctuates more in the other boroughs when traveling into Manhattan. Taxi congestion is also practically non-existent between 2 to 5 AM everywhere. This is also reflected in the time distribution graphs, where we see little to no pickups between the hours of 2 to 5 AM going into Manhattan from any borough. Little taxi traffic would suggest less people need to get around at that time, reducing the number of cars in the streets and traffic congestion.