Discovering Transit-Oriented Development Regions of Megacities Using Heterogeneous Urban Data

Abstract

Public transport is of great significance in megacities. Transit-oriented development (TOD) has become a reliable solution to urban sustainable development, which can reshape the urban form and improve its quality. This paper focuses on leveraging heterogeneous mega urban data to answer three critical questions in TOD: what region looks like under TOD concept, which regions have the potential to be TOD regions, and how to construct these TOD regions. For region partition, we propose a connected component-based clustering algorithm, which merges the large amount of public transport stops into representative cluster ones as region centers, and then apply the Voronoi algorithm to locate the region boundaries according to the cluster centers. For TOD region identification, we present a link importance-based random walk method that considers the importance of various transits and further identifies the most valuable regions to be TOD. For discovering functions of TOD regions, we introduce a multifactor-based function characterization approach that combines both the static linguistic factor and human mobility factor together and then derives the actual function distributions. The experiments, which are conducted on three real data sets, show the superiority of the proposed methods to solve the problems of region partition, TOD region identification, and function characterization for the megacities. In the meantime, the results provide support for the government to formulate public policy to construct a TOD city.