Urban Data Integration Using Proximity Relationship Learning for Design, Management, and Operations of Sustainable Urban Systems

Abstract

The world is rapidly urbanizing, with 66% of the world’s population expected to reside in cities by 2050. This massive influx of new urban citizens is putting enormous pressure on city systems and bringing forth challenges at the intersection of urban infrastructure, governance, and the environment. As a result, researchers and practitioners have turned to new advanced sensing and data analytics developed under the burgeoning smart city movement to improve the design, management, and operations of urban systems. However, it has been challenging to integrate, organize, and analyze the data emerging from urban systems due to their natural spatial, temporal and typological heterogeneity. This paper introduces an urban data integration (UDI) framework that is capable of integrating heterogeneous urban data. The proposed UDI framework is extensible to multiple types of urban systems, scalable to the growing volume of data streams (as a result of increasing geographical areas, higher sampling frequencies, and so on), and interpretable enough to help inform municipal decision-making. The UDI framework uses a series of proximity relationship learning algorithms to reconstruct urban data in a graph database. The merits, applicability, and efficacy of the proposed framework is demonstrated by validating and testing it on data from a midsize city in the United States and by benchmarking its interpretability and computational performance for a typical urban analytics scenario against current practice (i.e., a relational database). Results indicate that the UDI framework provides easier and more computationally efficient exploration and querying of urban data, and in turn can enable new computational approaches to urban system design, management, and operations.