Variable Coefficient Spatial Autoregression Model And Estimation

The traffic state of city has significant spatial characteristics,to explore the spatial characteristics of city traffic data.Use the traffic data of Guiyang city as an example.Firstly,Analysis the spatial correlation of Guiyang city traffic flow data.Based on spatial weight matrix by Moran 's I index to judge the autocorrelation of traffic flow.To construct spatial autoregressive model and the maximum likelihood estimation methods and then compared with the ordinary linear regression model,the results show that the spatial autoregressive model is better than the ordinary linear regression model.Secondly,Because of the spatial heterogeneity of traffic flow using kernel density estimation method,make a research on the distribution trend of different periods of the traffic flow,in order to discover the city road traffic flow of hot and cold regions.Based on the spatial heterogeneity of traffic flow to construct the geographic weighted regression model and investigate the traffic flow of spatial nonstationarity.Compared with the ordinary linear regression model,geographically weighted regression model shows obvious advantages.The prediction results of geographically weighted regression model visualized in the city of Guiyang on the map by ArcGIS software.Finally,According to the spatial correlation and spatial heterogeneity of traffic data,the spatial autoregressive model and geographically weighted regression model are extended to variable coefficient spatial autoregressive model.Using the non parameter B spline estimation method and get better estimation results,large sample properties are investigated,and simulations are performed to verify the estimation result.