Inferring demographics from human trajectories and geographical context
发布时间：2019-09-17
点击次数：
影响因子：
0.0
发表刊物：
Computers, Environment and Urban Systems
关键字：
Big geo-data; Demographics inferring; Human mobility; Trajectory mining
摘要：
The advances of positioning technologies and the widespread use of mobile devices bring us massive data with location information, or so-called big geo-data. One important part of big geo-data is massive digital human trajectories recorded by location-enabled mobile terminals and social apps. Digital human trajectories have been studied to learn more about human mobility and human activity. Existing research has shown that there exist strong associations between trajectory patterns and demographics. Given that demographics are essential information to various domains but not easy to acquire timely and on a large scale, inferring demographics from human trajectories has attracted attention from academia. In this paper, we proposed a demographics inferring framework suitable for big geo-data processing. Trajectory patterns were quantified from both spatiotemporal and semantic perspectives. Spatiotemporal features extracted from trajectories directly were used for capturing how people traveled in space and time. Semantic features obtained by attaching geographical context to tra-jectories were to reflect activities people conducted. Spatiotemporal and semantic features were organized into feature vectors and then input to supervised classification models to infer demographics. GPS trajectories and land use data in Beijing were used to validate the framework. Results show that the inference accuracies of marital status and residency status achieve 80% and thus prove the feasibility of our framework. This study can facilitate decision making in both business and social studies, such as personalized recommendation, commercial site selection and urban planning.
论文类型：
期刊论文
论文编号：
doi: 10.1016/j.compenvurbsys.2019.101368
学科门类：
理学
一级学科：
地理学
文献类型：
J
卷号：
77
页面范围：
101368
是否译文：
否
发表时间：
2019-07-30
收录刊物：
SCI
