Modeling Method Of Time Geographical Kernel Density Under Network Constraints

Time geography considers that the movement of moving objects between two space-time anchor points is uncertain,that is,the location of moving objects at any moment is uncertain.This uncertainty is constrained by the network in the urban road network space.The existing time geography method constructs a kernel density function under network constraints based on the principle of least effort,which realizes the expansion of traditional kernel density estimation in time geography,and expresses the effect that the visit probability of a moving object decays with the increase of the cost distance of the shortest path where the visit position is located.However,the above method ignores the effect that the visit probability decays with the increase of the minimum cost distance between the visit position and the anchor point,which follows the first law of geography.For this reason,this paper studies the kernel density estimation model that integrates the principle of least effort and the first law of geography in the road network space,and provides a new theoretical basis for the estimation of the visit probability of moving objects in the road network space.The main works of this paper are as follows:(1)This paper defines the reachable potential network area of moving objects,which is a carrier for the construction of the kernel density estimation model.On the basis of the principle of existing method,this paper introduces the first law of geography,deconstructs the saddle distribution characteristics of the Brownian bridge model in homogeneous space into double layer with the above principle and law,and then puts forward the conceptual framework of kernel density estimation in the potential network area based on the unification of the principle of least effort and the first law of geography.(2)According to this conceptual framework,this paper extends the existing method to construct the lower-layer visit density based on the principle of least effort,and establishes the upper-layer visit density based on first law of geography.On this basis,this paper realizes the mechanism of double-layer density co-location fusion,and then establishes a kernel density estimation logic model that takes into account the above two types of decay effects,which provides a theoretical basis for improving the density estimation of the road network space.(3)Based on the open vehicle trajectory data set provided by Didi Chuxing,two space-time anchor points are selected as typical examples,and the implementation and empirical analysis of the above logic model are carried out in this paper.The results show that the model proposed in this paper is effective in measuring the spatiotemporal uncertainty of the directional movement of vehicles in the road network space,and the comparison with existing methods and instance trajectories also shows the theoretical and practical significance of the proposed model containing saddle-shaped features.The research in this paper further improves the visit density estimation theory and method for time-geographical measurement of mobility uncertainty,and provides a basis for applications in scenarios such as visit intensity prediction and dynamic interaction of moving objects in the road network space.