Crowd-based urban characterization: extracting crowd behavioral patterns in urban areas from Twitter

The advent of location-based social networking sites provides an open sharing space of crowd-sourced lifelogs that can be regarded as a novel source to monitor massive crowds' lifestyles in the real world. In this paper, we challenge to analyze urban characteristics in terms of crowd behavior by utilizing the crowd lifelogs in urban area. In order to collect crowd behavioral data, we utilize Twitter where enormous numbers of geo-tagged crowd's micro lifelogs can be easily acquired. We model the crowd behavior on the social network sites as a feature, which will be used to derive crowd-based urban characteristics. Based on this crowd behavior feature, we analyze significant crowd behavioral patterns for extracting urban characteristics. In the experiment, we actually conduct the urban characterization over the crowd behavioral patterns using a large number of geo-tagged tweets found in Japan from Twitter and report a comparison result with map-based observation of cities as an evaluation.