Understanding Urban Dynamics from Massive Mobile Traffic Data.
Mingyang Zhang,Haohao Fu,Yong Li,Sheng Chen
DOI: https://doi.org/10.1109/tbdata.2017.2778721
2017-01-01
IEEE Transactions on Big Data
Abstract:Understanding the patterns of mobile data consumption is extremely valuable to reveal human activities and ecology in urban areas. This task is nontrivial in terms of three challenges: the complexity of mobile data consumption in large urban environment, the disturbance of abnormal events, and lack of prior knowledge for urban traffic patterns. We propose a novel approach to design a powerful system that consists of three subsystems: time series decomposing of mobile traffic data, extracting patterns from different components of the original traffic, and detecting anomalous events from noises. Our investigation involving the mobile traffic records of 6,400 cellular towers in Shanghai reveals three important observations. First, among all the 6,400 cellular towers, we identify five daily patterns corresponding to different human daily activity patterns. Second, we find that two natural patterns can be extracted from the weekly trend of mobile traffic consumption, which reflects modes of human activities. Last but not least, besides the regular patterns, we investigate how irregular activities affect mobile traffic consumption, and exploit this knowledge to successfully detect unusual events like concerts and soccer matches. Our proposed methodology therefore will aid a comprehensive understanding of large-scale mobile traffic consumption in urban areas.