Public Events Detection And Correlation Mining From Multiple Data Sources

Public security indicating the management level of a specific region is one of the most vital aspects in modern human civilization.Recently,public security incidents endanger the private properties and lives of citizens because of the enlarging gap of domestic incomes and a turbulent political situation around China.Meanwhile,the transmitting behaviors of incident messages indicate some complicated patterns regarding the prevalence of Internet technologies and social networks,which brings about challenges in traditional management of public security mechanisms.In this paper,we propose a novel,multisourcing data analyzing techniques for events detection and correlation mining of public security incidents,including intra-incidents transmission patterns and inter-incidents generality analysis on detected events.Our method is based upon the public,easily-accessed news and micro-blog data,in association with multiple data sources such as population and GDP distributions.We first finish the model construction of news reports and events based on geographic name tree and vector space model.Then we identify and extract different incidents from raw media data by event tagging based on context filtering and clustering algorithm based on Single-Pass,manually label the incidents by crowd source based on open data,and then the transmission patterns of intra-incidents and the generality of inter-incidents are analyzed with correlation analysis and visualization techniques.The research lay a solid foundation for the prediction of public security incident occurrence over temporal and spatial future,which indicates potential applications of our approach in the intelligent management of public security.