Research And Application Of Spatio-Temporal Analysis Method Based On INLA

The development of spatiotemporal analysis methods has provided more possibilities for processing complex spatio-temporal data sets and constructing complex spatio-temporal models.In the meantime,as Bayesian approximate computational methods are enriched,the practice of analyzing spatio-temporal data in a Bayesian framew ork is also increasing.Based on the theory of spatio-temporal point processes,this research assumes the urban management cases as spatio-temporal point patterns and regards the generation of urban management cases as inhomogeneous Poisson point processes.The data come from accumulated in the digital urban management system of P district in Northwest China.Under the Bayesian framework,this research used integrated nested Laplace approximations(INLA)and stochastic partial differential equations(SPDE)approach methods to modeling urban management cases as the spatio-temporal LGCP model.The research comprehensively analyzes the spatio-temporal distribution patterns and evolutionary trends of three types of urban management cases: street order cases,urban environment,and publicity advertising,then make predictions and warning of spatialtemporal distribution patterns in the next time knots.At the same time,the impact of urban built environment factors on urban management cases is also considered through Point of Interest(POI)data of cites.The study found that the spatio-temporal pattern of single urban management ceses in different regions is very different,and the distribution of different types of urban management cases in the same region is also quite different,spatio-temporal distribution of urban management cases are repeat and near-repeat happen where potential targets close to an initial incident are at a heightened risk of being actual targets.Considered the impact of the urban built environment factors on the urban management cases,a similar phenomenon also exists: different types of POI have different effects on the same type of urban management cases,as well as the same type of POI affects different types of urban management cases the impact of the incident is also different.These special differentiation regular patterns indicate that the occurrence of urban management cases and the formation of its spatial-temporal distribution has been profoundly affected by the specific urban environment,and it is the objective result of a combination of various factors.The study also compared the spatio-temporal LGCP model with the DBSCAN clustering method and kernel density estimation method.The experimental results show that the spatio-temporal LGCP model based on INLA and SPDE has a more excellent comprehensive performance in the extraction of the regular pattern of the spatio-temporal distribution pattern of urban management cases and the prediction of future evolutionary trends.The results of this research can provide scientific decision support for the urban management workflow of the urban management department,and also could improve the efficiency of the allocation of management resources of the urban management department,where is the excellent significance for strengthening the governance capacity of modern urban management and promoting good governance in cities.