Modelling of urban activity centers in Tokyo from alternative indicators of urban activity

Urban activity centers (UAC) – areas of the local concentrations of activities – is an effective spatial analytical tool for uncovering the spatial structure of cities. The problem is that due to the lack of the corresponding data, UAC were detected and analyzed by alternative indicators – the ones related to urban activities rather presumably than evidently. In this study, we build models allowing to predict the location of the UAC in cities based on such alternative indicators. We use the Person Trip Survey results for Tokyo to extract the indicator representing urban activities directly. Alternative indicators used in models as predictors are extracted from Geotagged tweets (Twitter data), data on the total floor area of buildings, and Points of Interest of OpenStreetMap. The modelling was conducted by regression and classification approaches involving the use of spatial autoregression. Models revealed that the identification of UAC by estimating the urban activity density using regression in which alternative indicators are predictors is a better approach than classifying UAC from alternative indicators directly. Differences between models' outcomes and their errors are discussed from the perspective of the nature of UAC in Tokyo and the specifics of the data used.