How representative are the geotagged tweets as a proxy for human mobility? Comparing social media data with travel survey and GPS log data

Abstract

Using geotagged social media data to model travel behaviour (human mobility) has received increasing attention. The low cost of retrieving geotagged tweets makes it especially appealing compared to other data sources. As a proxy for human mobility, such a data source needs careful inspection regarding its representativeness. To date, the works comparing geotagged tweets with the other data sources are not systematic; they either focus on individual trajectory or places network, however, the links between these two perspectives have rarely been constructed. Moreover, the involved data sources are usually limited to one to two categories without careful consideration of different data forms that support different research purposes. To fill the gaps in the literature, this paper attempts to conduct a systematic comparison, concerning longitudinal and lateral data forms where we combine individual trajectory and places network perspectives. This study is based on five data sets; they come from the household travel survey, geotagged tweets, and mobile phone GPS log. The results of this paper contribute a deeper understanding of what can be revealed from geotagged tweets compared with the other data sources to represent the actual travel behaviour.