Early Warning of City-Scale Unusual Social Event on Public Transportation Smartcard Data

Haiyang Wang, Xiaming Chen, Siwei Qiang, Honglun Zhang, Yongkun Wang, Jianyong Shi, Yaohui Jin

Abstract: A sudden social crowd event is serious to public safety as it usually triggers huge number of people who are overwhelming to existing public facilities. Detection, early warning of such social crowd events is very important to the city administration but a very challenging problem in research. In this paper, we aim to solve this problem by using the non-sensitive data from public transportation smartcard. We make a detailed analysis of the traffic of people from smartcard data,, we find a 'two-peak' pattern of human flow before, after a social crowd event happening. Motivated by this finding, we propose a framework for early detection of unusual social crowd events by exploiting time series analysis, machine learning technology. We evaluate our model on the real world public transportation data of the biggest metropolitan of China, validate our model with social crowd event data retrieved from the Internet. The evaluation result shows the effectiveness of our model.