Exploring the Potential of Uber Movement Data

Abstract

With the increasing use of big data in varied applications to improve decision making and provide new insights, the research explores the potential of the Uber Movement data set released by Uber comprising of travel times from one zone to the other. A better understanding of the potential of the dataset could lead to the addition of the existing tool kit of Transport planners and city officials at the municipality of Amsterdam. Moreover, it would be the first of a kind data set enabling an understanding of taxi movement in the city. The Uber Movement Travel Time comprises of the average travel time between two wijken, where the ‘sourceid’ and ‘dstid’ do not correspond to the origin and destination of a trip but simply represent the directionality of the travel time measured. The data is aggregated across different levels of temporal detail and the number of data points directly corresponds to the level of temporal aggregation. For instance, if the quarterly aggregated data for the different days of the week is downloaded, the number of data points between a ‘sourceid’ and ‘dstid’ cannot exceed seven.
Three aspects of the data set were explored: 1) ability to capture the demand for Ubers 2) ability to capture recurrent congestion and 3) ability to capture non-recurrent congestion. While the data according to the Uber Movement and previously used instances, the data is suited for performance (recurrent congestion and non-recurrent congestion) and impact-related studies of the network. The absence of route related information limits the applications of the data. The potential of the data is also limited by the data sparsity. The potential of the data was best revealed through demand studies which indicated a skewed user group of tourists, airport users (to and fro), work-related trips and users using Ubers late at night. In addition, with respect to the goals of the municipality in managing traffic activity across different zones and time periods, by implementing and extending an existing model in the form of adding ‘occupancy related measures’ and ‘shortest path’. Thus, based on the data penetration levels and travel time data, the model developed offers insights at a strategic level to the city in the form of Spatio-temporal concentration of Uber vehicles, occupancy levels through the day. The potential of the data lies in its ability to offer strategic insights to the city of Amsterdam and the greater Amsterdam region in the form of the unique Spatio-temporal spread of Uber vehicles across different hours of the day.