Spatiotemporal Pattern Analysis of Taxi Trips in New York City

A growing number of extensive data sets provides transportation planners with the necessary means to analyze urban travel patterns and gain insight into urban dynamics. This paper explores the spatial and temporal variation of taxi trips in New York City by analyzing 29 million trip records from a freely available data set. The study examined the role of airports in trip generation and attraction, as well as the variation of travel speed during the day. Comparison of hourly trip frequencies between weekdays and weekend days in each district revealed similarities and differences in the functional drivers of taxi trip demand. This paper presents a negative binomial regression model that predicts the number of taxi trips per district from the subway, train, and bus infrastructure, as well as socioeconomic and land use variables, where eigenvector spatial filtering is applied to explicitly model spatial auto correlation. Independent of the predictor variables, a combination of subway ridership and taxi trip numbers for each district in a mode mix variable allows, through the use of local indicators of spatial association statistics, the identification of districts that exhibit an increased inclination toward taxi use and that are currently poorly served by public transit. This approach could be used as a decision support tool to determine where investments in rapid transit infrastructure and service would be particularly effective to increase transit mode share.