Analysis and prediction of New York City taxi and Uber demands

Taxi and Uber are imperative transportation modes in New York City (NYC). This paper investigates the spatiotemporal distribution of pick-ups of medallion taxis (yellow), Street Hail Livery Service taxis (green), and Uber services in NYC, within the five boroughs: Brooklyn, the Bronx, Manhattan, Queens, and Staten Island. Regression models and machine learning algorithms such as XGboost and random forest are used to predict the ridership of taxis and Uber dataset combined in NYC, given a time window of one-hour and locations within zip-code areas. The dataset consists of over 90 million trips within the period April-September 2014, yellow with 86% the most used in the city, followed by green with 9%, and Uber with 5%. In the outer boroughs, the number of pick-ups is 12.9 million (14%), while 77.9 million (86%) were made in Manhattan only. Yellow is the predominant option in Manhattan and Queens, while green is preferred in Brooklyn and Bronx. In Staten Island, the market is shared between the three services. However, Uber presents a highly rising trend of 81% in Manhattan and 145% in outer boroughs during the analysis period. The regression model XGboost performed best because of its exceptional capacity to catch complex feature dependencies. The XGboost model accomplished an estimation of 38.51 for RMSE and 0.97 for R^2. This model could present valuable insights to taxi companies, decision-makers, and city planners in responding to questions, e.g., how to situate taxis where they are required, understand how ridership shifts over time, and the total number of taxis needed to dispatch to meet de the demand.