Research On Multi-Dimensional Analysis And Mining Of Urban Human Mobility Pattern

Travel behaviors of urban residents are important component parts in urban mobility system.The analysis on the characteristics of urban residents' mobility patterns is helpful for scientific urban planning and traffic management.With the application and development of technologies such as mobile communications,positioning and storage,researchers have been able to obtain a large amount of urban residents' travel data.How to mine the hidden mobility patterns of residents,useful information and knowledge from these high-dimensional and complex travel data is a hot and difficult point of current research.Based on the multi-dimensional characteristics of urban residents' travel data,this paper combines statistical methods,complex network theory,spatio-temporal data mining methods,machine learning and other research methods to establish some new calculation models and analysis methods,from three levels of "single dimension-spatiotemporal dimensions-multi-dimensional " to study mobility patterns of urban residents under different analysis dimensions.The main research contents and conclusions of this paper are as follows:(1)Through the empirical research on the two datasets of New York City taxis and public bicycles that characterize the long-distance and short-distance trajectories of urban residents,the analysis reflects the three movement indicators of the travel patterns of different travel groups in the time and space dimensions: The travel distance,travel time,and community network migration at higher spatial scales.It is found that different travel modes and different groups of people generally have the following characteristics in a specific city:(1)The distribution of travel distance and travel time obeys log-normal distribution;(2)The daily travel of urban residents has a specific time budget(the upper limit of 30 minutes for New York City residents);(3)Urban residents have specific travel time and travel distance preferences when choosing different modes of travel;(4)Due to the combined effect of various factors such as administrative division,geographical proximity,travel purpose and travel distance,the travel activities of different groups in the city have specific time and space agglomeration and spatial limitations.The above empirical results provide evidence that the travel behaviors of different travel groups in a particular city have the same internal mechanism.(2)As the traditional spatiotemporal clustering method does not fully consider the spatio-temporal correlation,a two-step strategy for spatiotemporal co-clustering travel pattern mining method is proposed,which realizes the mining of spatiotemporal patterns with different spatiotemporal granularities.This method first uses the coclustering algorithm to divide the spatio-temporal co-occurrence matrix into smaller homogeneous co-clustering clusters,and then uses the K-means algorithm to gradually group the adjacent clusters of the co-clustering clusters.Intuitive visualization and experimental results show that this method improves the accuracy and stability of the clustering results.Finally,a case study was carried out using GPS data of taxis in New York City,and the practicability of the method was verified by extracting the spatiotemporal interaction pattern,which provided an auxiliary decision-making basis for taxi companies to implement vehicle scheduling.(3)From the perspective of human-land coupling,a multi-dimensional travel pattern analysis model is constructed,which integrates the static attributes of the crowd into the temporal-spatial relationship model,and promotes the identification of activities of different groups.The model application analysis based on New York City bicycle rental data verified the effectiveness of the model,and found the intrinsic correlation among the user's travel time,space pattern,the user's age attribute and travel distance,identifying specific areas where elderly users are intensive in cycling activities during daytime hours,as well as areas where young users are riding intensively at night.This article also found some rules of time and space for bicycle travel:(1)The regular travel distance of 1?3km is the main distance that urban residents choose to travel by bicycle;(2)Young users tend to choose long-distance cycling activities during the peak hours of workday evenings,and older users tend to travel in in the morning peak and daytime,and the distance of the trip is short.The unique travel patterns of these different groups of people provide a scientific basis for transportation infrastructure planning and investment.