Urban function connectivity: Characterisation of functional urban streets with social media check-in data

Abstract

Social media check-in data, one type of crowdsourcing open data about individual activity-related choices, provides a new perspective to sense people's spatial and temporal preference in urban places. In this paper, through the analysis of the interaction between these scored places on streets, we aim to advance our knowledge of network accessibility with social media check-ins to portray urban structure and related socioeconomic performance more explicitly. By conceptualising an interface graph to reflect the interplay between land-use points and the co-visual paths, we propose a novel framework to characterise the urban streets with land-use connectivity indices that are measured with a new type of place-function signature. A “3-Ds” model is introduced to package three principal dimensions of urban function network, including accessible density, accessible diversity and delivery efficiency, as one integrated index that works towards a comprehensive understanding of function connectivity from each street's midpoints to all reachable land-use points. Streets are further partitioned to the annotated function regions based on function connectivity in different types of active land-use. The results of preliminary studies in the city of Tianjin, China show that the proposed metrics can explicitly describe the inherent function structure and the regions' typology across scales. Compared with space syntax measurements at the same radius for describing the variation of empirically observed house price, the integrated metric can improve the predictability of statistic models sufficiently, and each specified index is confirmed to be statistically significant by controlling other factors. Overall, this research shows that the usage of ubiquitous big social media data can enrich the current description of the urban network system and enhance the predictability of network accessibility on socioeconomic performance.