Beyond expected regularity of aggregate urban mobility: A case study of ridesourcing service

Abstract

The regularity of urban mobility is a tacit understanding in public transportation planning. Many studies investigated regularity in the context of individual travel while fewer were found on aggregate mobility. This work proposes a generic measure of regularity to quantify the degree of repetition of a time or spatial data series. In addition, it discusses how the properties of a data series itself may dictate the regularity and adopts parametric bootstrapping to estimate the residual regularity that deviates from realizations of stochastic processes. The proposed measure can examine not only the periodicity of trip generation in locations but also the stationarity of travel flows. The experiments were conducted on a ridesourcing dataset, including more than six million trips in Chengdu, China. Both the temporal and spatial regularities were investigated at distinct time intervals. We discussed how the total travel volume and the peakedness of probability distribution could affect the regularity. It was found that regularity was positively associated with the total volume and the peakedness. We further examined the impact of the built environment on regularity and the effects of spatial and temporal scales. The results show both point-of-interest density and diversity and tourist attractions contribute to regularity. Regularity grows linearly as the spatial and temporal scales increase exponentially. Lastly, different specifications of regularities were compared, and the outcome was generally consistent across different similarity measures including coefficient of determination, Manhattan distance, Euclidean distance, cross-entropy and cosine similarity.