Analyzing Distortion of Geo-social Proximity using Massive Crowd Moving Logs over Twitter

摘要

Due to the complicated transportation network associated with the progress of urbanization, a sense of proximity between places often seems to be distorted from physical proximity based on geographical distance. It is crucial to measure people's sense of proximity for public aims such as urban analysis and planning as well as personal aims such as neighborhood area search. We consider that personal lifelogs on recent social network services would become an important clue to grasp the sense of proximity. In this paper, we analyze the distortion of proximity between urban clusters by exploiting crowd experiences. For this, we utilize crowd travel logs from Twitter and person trip survey on a national survey, respectively. Next we measure crowd-based proximity between urban clusters and re-map the urban clusters on 2-dimensional space depending on their geo-social proximity by applying Multi-Dimensional Scaling (MDS). Then, we extract a socio-geographic structure of the urban clusters in the form of a graph represented by Minimum Spanning Tree (MST). In the experiment, with two different crowd-sourced datasets; geo-tagged tweets and the person trip survey, we observe the distortion of proximity in term of travel time and the amount of travels, respectively.