Research Of Spatial-temporal Clustering For Case Data Based On The Public Security Big-data

With the continuous construction of "Golden Shield Project" of China Police,the public security informationization has made a breakthrough in our country.Meanwhile,a variety of sources and types of social information pouring into public security systems.The Public Security Big-data is the core of the public security department's data resources and the foundation of the police activities.One of the most difficult challenges which the public security department is facing,is how to integrate and mining those data.This paper which considers GIS peculiar expression mechanism of space and time,launches the research on this issue with the perspective of data and application.Then,this paper explorers the space-temporal clustering algorithm for case data.Through the study of this article,police officers can find crime patterns,realize the efficient application of The Public Security Big-data and achieve reasonable allocation of police resources.The main content of this paper lists below:(1)The Public Security Big-data integration method driven by space and timeBased on the characteristics of massive,multi-sourced heterogeneous,Value High and Low Density,spatial-temporal of The police big data,this paper regards the space and time as the framework,Geographic entity as the core,relationship with elements as the clue and bulids the police geographic information integration framework and the spatial model.The experiment demonstrates that by the steps of data acquisition,tagging information,data loader,Spatio-temporal data modeling,The Public Security Big-data can be integrated together effectively.(2)Exploratory analysis of spatio-temporal case dataThe Exploratory analysis of spatio-temporal case data includes the analysis of spatio-temporal data correlation and heterogeneity,whitch is the key step of the space-temporal clustering algorithm.This paper research the spatial correlation by the global space Moran's I index,the temporal correlation by temporal auto-correlation function(ACF)and the spatio-temporal correlation by global spatio-temporal Moran's I index(STI).The experiment demonstrates that spatio-temporal events have strong correlation and stable tendency.(3)The methods of spatio-temporal clustering under the different conditionsBased on the two kinds of typical algorithm of ST-DBSCAN and STSNN,this paper explorer how to discover spatio-temporal clusters of case data under different conditions by building the spatio-temporal neighbors and spatio-temporal shared nearest neighbors.The experiment demonstrates that ST-DBSCAN algorithm is more suitable for the cases data whitch has rich types and Large time scale,while STSNN algorithm is efficiective for discovering the spatio-temporal clusters of cases data which have a single type and Small time scale.