Research On Mining Residents’ Trip Attractive Areas And Popular Routes Based On Taxi Trajectory Data

Residents’ travel behavior is an important reference basis for urban construction planning and comprehensive management of urban transportation.In recent years,with the continuous development of mobile positioning technology and location-based service technology,the acquisition of trajectory data has become more and more convenient.The trajectory data contains rich spatiotemporal semantic information.Mining and analyzing the trajectory data can extract the latent knowledge and patterns in the trajectory data.This paper takes Zhengzhou City as the research area,and takes taxi trajectory data and points of interest data as the main data sources.Extracting valid data from trajectory data based on trajectory preprocessing technology.Based on this data,a clustering algorithm based on data fields and CFSFDP is used to extract hotspot areas in the city,and a trajectory clustering method based on density core is used to extract hotspot paths in the city,analyzing residents’ travel laws in conjunction with urban functional areas divided by points of interest data,and it can provide an important reference for the city’s traffic management and urban construction.The main research results of this paper are as follows:(1)Aiming at the problems of difficult parameter selection and difficulty in determining the number of classes in traditional hot-spot clustering methods,a clustering algorithm based on data field and CFSFDP is proposed.The algorithm introduces a Gaussian function to calculate the potential of the data point,and at the same time combines the information entropy to select the impact factor,and automatically selects the clustering center through the "elbow" law.The algorithm in this paper selects manual test data and taxi trajectory data as experimental data,and compares with K-MEANS、DBSCAN、CFSFDP classic clustering methods to verify the effectiveness of the algorithm.Experiments show that the algorithm in this paper can identify clusters and noise points well,and can automatically select parameters,avoiding the uncertainty caused by manual parameter selection.(2)Aiming at the problems of low efficiency and poor scalability of massive taxi trajectory data faced by traditional trajectory clustering methods,this paper proposes a density-based taxi trajectory clustering algorithm.The algorithm first improves the SP similarity distance.By calculating the bidirectional distance of the trajectory,the accuracy of the similarity algorithm is improved,and it is suitable for taxi trajectory data of different lengths.Secondly,the algorithm stores detailed information of clusters by setting cluster nodes,and uses dense core trajectories in clusters to aggregate similar trajectories.The algorithm selects the taxi trajectory data for comparison and analysis with TRACLUS 、 OPTICS clustering algorithms.Experimental results show that the algorithm in this paper has good scalability and high execution efficiency,and can be well applied to trajectory clustering.