Research On Spatio-temporal Trajectory Data Mining And Its Application In Urban Residents' Travel Behavior Pattern Analysis

Activities and travel behaviors of urban residents are important component parts in urban activity-travel system.The analysis on the characteristics of urban residents' travel behaviors can help scientifically carry out urban planning and traffic management.However,obtaining information on residents' travel activities through traditional manual surveys and statistical analysis methods has not been able to meet the needs of rapid development of urbanization.On the other hand,with the application and development of information technologies such as communication,positioning and storage,a large number of mobile trajectory data traveled by urban residents can be collected and stored.These trajectory data contain rich spatio-temporal semantic information.A great deal of valuable travel information can be obtained based on the mining and analysis of trajectory data.Furthermore,the daily behavior patterns of individual users and the spatial distribution characteristics of group users' movements can be found.The results can effectively serve the fields of intelligent transportation and urban planning.This paper is devoted to the research on spatio-temporal trajectory data mining and its application in the analysis of urban residents' travel behavior patterns.First,the spatio-temporal trajectory data model is constructed,and a trajectory outlier detection method based on common slices sub-sequence is proposed.The abnormal routes are analyzed based on the individual travel trajectory data of Beijing residents.Second,a trajectory clustering method based on multi-feature similarity measure model is proposed.The method is further verified by the analysis of residents' regular travel routes.Then,a trajectory classification method based on information entropy and multi-layer BP neural network is proposed to identify transportation modes used in the travel routes of residents.Finally,based on the taxi trajectory dataset of Beijing,the spatial characteristics of urban residents' travel are analyzed from the perspectives of hotspot distribution and mobile mode.A hotspot location mining method based on neighborhood-associated quality clustering is proposed,and an example verification is carried out through residents' travel commuting analysis and taxi hotspot area detection.The main work and innovations are as follows:(1)Aiming at the difficulty in identifying local anomalous trajectories in residents' travel routes,a novel trajectory outlier detection method based on common slices sub-sequence is proposed.First,the trajectory direction-code sequence and trajectory slice sequence are extracted.The CSS distance between two trajectories is measured by identifying their common slices sub-sequences.Then slice outliers and trajectory outliers are detected based on the CSS distance calculation.Comparison experiments on the synthetic trajectory dataset with classification labels show that the proposed method is suitable for trajectory slice partition and trajectory outlier detection,and the detection rate is better than that of existing methods.Finally,the proposed method is applied to the real daily travel trajectory dataset provided by the GeoLife project to detect the abnormal travel trajectories of residents and analyze the possible causes.(2)Traditional trajectory clustering methods do not fully consider the internal and external features of trajectory data,and cannot effectively analyze the conventional routes of residents' travel.For this issue,a trajectory clustering method based on multi-feature similarity measure is proposed.The method uses the characteristics of trajectories including time,orientation,speed,shape,location and continuity to achieve the similarity measurement between trajectories,and optimizes the selection of initial trajectory centers based on the trajectory duration,which improves the accuracy and stability of the clustering results.Finally,the method is applied to the analysis on conventional routes of urban residents' travel,and its practicability is verified.Application and analysis results of this method provide an auxiliary decision-making basis for urban management and traffic planning.(3)Travel transportation mode is the main component of the characteristics of residents' travel behaviors.Taking the identification of residents' travel transportation modes as application background,a trajectory classification method based on information entropy and multi-layer BP neural network is proposed.This method comprehensively considers the geographic information including timestamp,latitude,longitude and road section to extract the motion features of each location.The features are filtered and weighted using information entropy.Then a multi-layer neural network structure is constructed for trajectory classification.Based on the post-processing mechanism with identification,the accuracy of transportation mode identification is improved and the identification results are optimized.Experimental results compared with traditional machine learning methods and related algorithms show that the classification method proposed in this paper has higher accuracy.It can effectively identify the transportation modes adopted by urban residents,make up for and solve the difficulties as well as shortcomings in traditional travel investigation.(4)Taking the Beijing taxi trajectory dataset as data source,this paper analyzes the spatio-temporal characteristics of urban residents' travel from multiple angles,and effectively discovers urban hotspots and regional movement patterns.First,the taxi travel trajectory and passenger location data are extracted,and an urban residents' travel hotspot detection algorithm based on neighborhood-associated quality clustering is proposed to extract effective and reasonable urban hotspots from the taxi travel trajectory data.Then,the experimental verification and visual analysis are carried out in terms of the spatial distribution characteristics of passenger locations,the regional movement patterns of taxi travel trajectories,and the distribution of hotspots for urban residents.The results show that in-depth mining results on taxi trajectory data contribute to the analysis on spatial characteristics of urban residents' travel.The trajectory hotspot area detection results can provide valuable services for the traffic management system.