Social Security Event Detection Based on Characteristic Subspace Learning
This paper uses network big data to detect social security events,and proposes an event detection method based on characteristic subspace learning to solve the problems of the existing methods,such as poor robustness of event characteristic subspace for interference characteristics and low subspace change tracking accuracy.Firstly,the detection of the initialized event cluster is carried out,the semantic network structure of key characteristics is extracted and the sample clustering is carried out,the robust principal component analysis is used to estimate the event vector subspace,and the online update mechanism of the event subspace is designed to track the semantic drift of the event-related network text with the advancement of time.Experimental verification shows that the proposed method can effectively detect social security events,realize the subspace estimation of low-dimensional characteristics of event clusters,and effectively track the development of event characteristics.
NETL
NSTL
万方数据
