Research On Clustering Method For Large-scale GPS Trajectory Data

In recent years,with the popularization and deep integration of mobile Internet,GIS(Geographic Information System),Internet of Things and other technologies in the transportation field,large-scale urban traffic trajectory data has been able to be effectively collected,transmitted,and stored.At the same time,driven by the rapid development of cloud computing and big data analysis technology,it is one of urgent needs to provide real-time and effective personalized travel services for traffic participants by efficient analysis and research on large-scale urban traffic trajectory data in the field of intelligent transportation.Trajectory clustering analysis is an important research content of large-scale urban traffic trajectory data analysis.By clustering large-scale urban traffic trajectory data,trajectory clustering analysis can extract hidden semantic information in urban traffic activities,excavate travel rule in urban traffic activities,discover travel hot spots or abnormal paths and provide effective decisions support for personalized travel services.Based on this,this paper researches the clustering algorithm for large-scale urban GPS trajectory data.The specific research works as follow:(1)A dynamic time warping algorithm(SDTW+)for time-segmentation and segmentation is proposed.This algorithm effectively improve the calculation accuracy by adjusting the shape factors of the trajectory.The experiment is combined with hierarchical clustering algorithm to achieve clustering.The results show that the clustering effect based on the algorithm in this paper is significantly improved compared to the two clustering algorithms based on SDTW and DTW(2)A parallel trajectory clustering method for large-scale GPS trajectory data is proposed.This method divides the data into different regions through the grid,and proposes a grid edge virtualization scheme to ensure the correct clustering of the edge data;by parallel design and implementation of the DBSCAN clustering algorithm based on SDTW +,it satisfies the clustering of large-scale trajectory data Claim.The analysis of the experimental results show that the algorithm has good performance in terms of scale growth,acceleration ratio and scalability.(3)A LCS hot-paths extraction algorithm based on segmented metrics is proposed.This algorithm improves the common sub-trajectory matching scheme and adds thermal values to extract hot-paths.The experimental results show that of this algorithm can effectively extract hot-paths based on trajectory clustering.