Neighborhood Association Modeling And Its Application

Spatio-temporal analysis and modeling has always been a research hotspot in the field of geography.Time and space are two basic dimensions for studying the geographic process changes under the influence of human activities and natural factors.The association beween spatial objects in spatio-temporal dimensions is the prerequisite for exploring the formation and development of various spatio-temporal patterns.Spatial neighborhood and spatio-temporal neighborhood are important factors in the study of mining spatial associations under the background of geographic spatio-temporal research.Spatial neighborhood reflects a specific spatio-temporal range within which spatial objects may be related with other objects in the process of spatio-temporal change,and represents the spatio-temporal attribute and structure in the spatial scope,and may produce impacts on the occurrence and development of geographic events.The study of neighborhood association modeling has important implications for analysing and predicting the occurrence and development of geographic events.For the current methods of studying neighborhood association modeling based on spatio-temporal association patterns,the neighborhood effect of spatio-temporal association is not fully considered,and thus the spatio-temporal multidimensional associations are not correctly discovered.The outcome of the issue limits the neighborhood association study in spatio-temporal association models,possibly leading unsuccessful analysis and prediction of the occurrence and development of geographic events.Due to the above problems,this thesis focuses on the theory and method of neighborhood association modeling.The study tries to establish the theory model of neighborhood association for mining local spatial and spatio-temporal association patterns,proposes the algorithms for the model,and examines the role of neighborhood association effect in geographic process analysis.According to different types of geographic research objectives,the models of spatial neighborhood,spatio-temporal neighborhood and neighborhood rectified geographically weighted regression are established,and the algorithms of neighborhood association effect are proposed.The proposed neighborhood association model and algorithms are applied to typical geographic fields to verify the importance of neighborhood association effect in spatio-temporal association analysis.The research contents and research achievements include the following aspects:(1)Due to the incomplete consideration of neighborhood association effect in neighborhood modeling process,to model the spatial neighborhood association,this thesis proposes the basic idea to quantify the spatial neighborhood association effect,gives the quantitative index of spatial neighborhood association,constructs the spatial neighborhood association model,and implements the algorithm for the spatial neighborhood association model.The spatial neighborhood association effect is synthetically calculated from the neighborhood spatial scale,neighborhood kernel structure and other aspects,which improves the performance of spatial neighborhood association model significantly.(2)Due to the lack of time series association in the process of neighborhood modeling,this study introduces time scale to expand the connotation of spatial neighborhood association modeling based on the spatial neighborhood association model.The expression method and model of spatio-temporal neighborhood association are proposed.The measurement index of spatio-temporal neighborhood association is constructed,the corresponding algorithm of spatio-temporal neighborhood association model is given,and the function of spatio-temporal neighborhood association model in the process of geographical spatio-temporal change is verified.The above method is used to quantify the influence of different neighborhood spatial scales and different neighborhood time windows in neighborhood association modeling analysis.(3)Due to the random errors of independent samples in classical geographically weighted regression,this study proposes a neighborhood rectified geographically weighted regression method considering neighborhood association in samples selection based on geographically weighted regression,and gives the algorithm of the model,which improves the stability of geographically weighted regression model and fits for continuous data neighborhood association modeling.Combined with spatial neighborhood association and spatio-temporal neighborhood association modeling,the theory and method of neighborhood association modeling are illustrated.(4)The proposed neighborhood association modeling theory and algorithms are applied to typical geographic fields,including vegetation degradation and restoration,multi-factor association analysis in grassland ecosystem,simulation and prediction of typical urban expansion process.In summary,this thesis gives a general method to analyze neighborhood association effect and neighborhood modeling in typical geographic process,and provides decision support for analyzing,understanding and regulating geographic process.