423 / 2019-02-27 16:12:38

Virtual activity and physical mobility: an exploratory investigation

Introduction
The past decades have witnessed a great improvement of information and communications technology (ICT) in many countries. There is no doubt that to some extent the time spent in the virtual space, where people obtain information, extend social networks, and stay entertained, etc. has shaped their mobility patterns in physical space. Based on individual level cell phone data with aggregated records of virtual and physical activities, this paper focuses on two questions, 1) what is the features that characterize the virtual activity, and 2) is there any association between virtual activity and physical mobility and, if yes, what implication does it have with respect to theories and hypothesis in time geography.
Data and methods
Our dataset consists of anonymous records collected from 640 thousand mobile phone users with both their virtual activity via APPs and physical activity within Shanghai from 12/27/2015 through 01/06/2016. To avoid any possibility of privacy disclosure, the operation company aggregated the spatial locations of physical activities every one hour and grouped the virtual activities by APP types beforehand.
Given this dataset, we utilized two dimensions to describe the virtual activity, i.e., the number and types of visits a user conducted during the period. Three aspects are analyzed: the population distribution of the web visits, the heterogeneity between APP-types, and the correlation among them.
After that, we associated virtual activity with physical mobility, for which we focused our interest on the spatial distribution of residence/workplace and their daily mobility. First, several subsamples of active users are extracted and labeled by their web visits. Then, by mapping the distribution of their residence/workplace, we investigated the relations between the virtual group labels and the locations of their residence/workplace and concluded potential explanations combining the background of the urban configuration. Furthermore, we tested several hypotheses about the correlation between virtual activity and commuting cost as well as the numbers/range of non-commuting activities.
Conclusion and discussions
We showed that both the overall distribution of web visit numbers of individual users and of virtual activity types are typical zeta distribution. And when a user spends more time in virtual space, his/her interest will also concentrate more significantly in primary preferred virtual activity type. We also found that the preference of some virtual activity types are correlated with a declining trend from the most popular type to the least.
Additionally, we provided clear evidence that virtual activity is associated with physical mobility. Specifically, the inner city has a comparatively low web visits than the outer region. The most active virtual visits come from the CBD area and the university towns. The result may be explained by the spatial configuration of urban function and demographics.
Furthermore, we found that when web visit grows both the average commuting distance and the time/range of non-commuting activities first increases then goes down. The beginning increases could be attributed to the inactive users, e.g., the elders, while the latter decreases imply a significant suppression on physical mobility produced by excess virtual activities.
Our results show great potential for improving the theories in time geography and the understanding of future human mobility in both virtual and physical space.