The Research On Anisotropic Geographically Weighted Regression Model And Its Application In Mineral Exploration

In the comparative analysis of spatial data,the geographically weighted regression(GWR)model is more and more widely used due to its advantages in spatial analysis.As a new spatial analysis method,the GWR model explores the spatial non-stationarity of the relationship between the ore-controlling elements and mineralization by introducing spatial locations.With the enhancement of theoretical understanding of geology,more and more geologists realize that Mineralization is a complex geological process.The spatial characteristics of the mineralization process and its products often show anisotropy.However,the weight kernel function used in the traditional geographic weighting model has isotropic characteristics.From the perspective of the algorithm itself and the nature of mineralization events,the weight kernel function cannot objectively and truly reflect the complex geological processes with anisotropic characteristics and the spatial distribution of their products.In addition,variables or data in various fields such as social economics and agriculture are spatially anisotropic.GWR can no longer satisfy the precise description of the relationship between their variables.Based on the above problems,this paper improves the isotropic weight kernel function and constructs a new geographic weighted regression model based on the anisotropic kernel function.This study applies the traditional geographic weighted regression model and the improved geographic weighted regression model to quantitatively evaluate the spatial instability of the ore-controlling factors in the east Tianshan mining area.By comparative analysis,the effectiveness,feasibility and advantages of the improved Geographically Weighted Regression Model is verified.The main research contents and results are as follows:(1)Improvement on the weight kernel function and bandwidthConsidering the anisotropy of spatial variables,the circular bandwidth in the traditional GWR model is improved to an elliptical bandwidth.The ellipse is made directional,and used to divide the spatial scope of the regression analysis.This paper improves the isotropic weight kernel function in traditional GWR to a new anisotropic weight kernel function algorithm.(2)Evaluation method of regression modelThrough the research on the evaluation of the traditional GWR model,it is found that the current evaluation parameters include goodness of fit,Moran's Index,and AIC criteria.Through studying its basic theory and algorithm principle,this paper achieves the transplantation of the evaluation parameters of the anisotropic weight kernel function.(3)Application demonstration of geographic weighted regression model based on anisotropic weight kernel functionTaking the East Tianshan metallogenic belt in Xinjiang as the study area,based on the marine volcanic sedimentary iron deposit metallogenic model in the study area,an application demonstration study of the improved geographic weighted regression model based on anisotropic weighted core function was carried out.Compared with the traditional GWR model,the improved GWR is more advantageous in exploring the non-stationarity of the spatial relationship.It can more quantitatively display the spatial non-stationarity and anisotropy of various ore-controlling elements on the metallogenic control.