Visual Analysis System Based On Trajectory Triple Feature Clustering

With the advent of the Internet of Everything,people’s travel trajectories have become more and more accessible,and the research on people’s travel trajectories has become more and more in-depth.With the deepening of research,people gradually discover the value information hidden behind the trajectory,from the trajectory of people movement to the trajectory of vehicle travel,all of which reflect the laws of human society’s movement.However,the amount of trajectory data is large and the value density is low.To explore its hidden value information,it is necessary to analyze and process the trajectory data.Trajectory clustering has always been a common method used by researchers.Through clustering,the regular characteristics and behavior patterns of trajectories can be discovered.In recent years,the trajectory clustering method has developed from a single spatial clustering to a clustering based on time and space.Compared with spatial clustering,spatio-temporal clustering can subdivide trajectories of the same spatial location according to time,making the clustering logic more rigorous and the clustering results more reliable.However,trajectory data includes not only space and time information,but also various attribute information,such as: driving speed,driving direction,passenger status,etc.If the attribute information is ignored in the clustering process,the clustering will be incomplete and the results will be inaccurate.Problems such as low application value.In order to solve these problems,this thesis has done the following two aspects:1.Propose a triple feature clustering method based on trajectory space,time and attributes.This method adds attributes to the constraints of trajectory clustering.The main process of the algorithm is: first preprocess the trajectory data,and treat the processed trajectory data as sentence text,and then use Doc2 Vec to convert the sentence text into high-dimensional vectors.Finally,K-Means is used to cluster high-dimensional vectors,so as to achieve triple feature clustering of trajectories.2.Use visualization technology to display clustering results and form a visual analysis system.The system is mainly divided into three parts: trajectory diagram,attribute diagram and word cloud diagram,which respectively display the space,time and attribute information of the trajectory.Through the interactive function,the three parts can be linked and analyzed,and the trajectory after clustering can be observed the regular characteristics in the three dimensions of space,time and attributes after clustering.Finally,it was verified with Shenzhen taxi trajectory data.Experiments show that this method analyzes the behavior patterns and travel characteristics of the vehicles by observing the spatial distribution,time distribution and attribute characteristics of the taxi trajectory in Shenzhen,which has certain practical application value.Compared with the existing spatio-temporal clustering,the clustering method proposed in this thesis can make the clustering logic more rigorous,the clustering results more refined,and the travel characteristics more obvious after adding attribute constraints.