Exploring the Evolutionary Patterns of Urban Activity Areas Based on Origin-Destination Data

An understanding of the evolutionary patterns in areas of urban activity is crucial for official decision makers and urban planners. The origin-destination (OD) datasets generated by human daily travel behavior reflect urban dynamics. Previous spatio-temporal analysis methods utilize these datasets to extract popular city areas, with the ignorance of the flow relationships between areas. Several methods have been unable to determine time steps with similar spatial characteristics automatically or failed to recognize the evolutionary patterns of various modalities for a city. In this paper, we propose a new methodology to discover the hidden semantic-level city dynamics from OD data. The method carries out spatial simplification and constructs a sequence of location networks at first. Then, the hourly network is studied as a document consisting of trip relationships among location clusters, enabling a semantic analysis of the OD dataset as a document corpus. Hidden themes, namely, traffic topics, are identified through a topic modeling technique in an unsupervised manner. Finally, an interactive visual analytics system is designed to intuitively demonstrate the probability-based thematic information and the evolutionary activity patterns of a city. The feasibility and validity of our method are demonstrated via case studies with two kinds of real-world datasets: bike-sharing system (BSS) dataset and taxi dataset. Semantic-level city structures and recurrent behaviors representing the life of a large set of users, as well as the differences in BSS usage patterns of two cities are discovered. We also discover how people use different means of transportation for one city.