Research On The Identification Method And Behavior Complexity Of Urban Rail Transit Group Trips From The Perspective Of Big Data

Traditional travel modeling is generally based on individual decision-making,while group travel reflects the possible social relationships behind the spatial trajectory.Existing research focuses on individual mobility modeling as well as the identification of individual travel behavior patterns while ignoring how social interactions among individuals affect mobility in complex behaviors such as group travel.As a typical behavior of inter-personal relationships in travel,the study of group travel patterns in urban settings can help to understand the social relationships and behavioral characteristics of urban residents.Group travel behavior facilitates research areas more than traffic demand management and forecasting.Since group travel reflects a stronger inter-personal connection,by analyzing the group travel flow and travel purposes,it is beneficial to analyze how the flow carried by people spread dynamically in the network.Relative spreading processes include infectious viruses and word-of-mouth information,which are both closely related to the intensity of contact or exposure.Therefore,identifying social networks from spatio-temporal trajectories contributes to the study of complex travel behavior.Quantifying group structure and proportion is also of concern to emergency management.For example,when dispatching bus shuttles for evacuation during special events,passengers traveling together will form one single decisionmaking unit instead of multiple decision-making groups.In addition,under emergency,the group travel members will consider other partners when making an escape decision.Therefore,the number of members in each group and the strength of the connection will impact the overall decision-making.In addition,the purpose of activities,the proportion,and the structure of the group network can also be used as a reference for the emergency evacuation in the destination locations outside the metro station.Although group travel behavior is of great significance for travel planning,information dissemination,and emergency management,the lack of large-scale individual offline interaction information data makes such social relationships and the interactions between social relationships and travel remain poorly understood.Based on a large data set of more than 120 million trips in the Beijing metro network,this paper proposes a method to identify group travels from familiar strangers formed by overlapping travel trajectories from the massive trajectory data of metro passengers.Through the identification of group travel,the social relationship implied in human activities in the city is identified in a larger space and time range.Based on the social relationship formed by the extracted group travel passengers,the characteristics of travel and social network and the interaction between them are analyzed.Combined with the population mobility network,the dynamic propagation process on the social travel network is analyzed.The main research contents of this paper are as follows:(1)The temporal and spatial distribution characteristics of individual travel from the micro-perspective and the collective travel from the macro-perspective are studied respectively.Based on the large-scale AFC data,the daily travel patterns and preferences of passengers were identified,clustered,and modeled.For individual temporal patterns,we study the fluctuation of trip demand for passengers with different trip counts,the inter-trip interval,and the stay time distribution of consecutive trips.For individual spatial patterns,spatial diversity,the static visit pattern,as well as the form and spatio-temporal characteristics of consecutive trips are analyzed.For collective temporal patterns,based on the fluctuation of inflow and outflow,metro stations are clustered to find the functional areas covered by the stations;the temporal variability of inflow,outflow volume,and OD volume are analyzed to give implications for prediction.For collective spatial patterns,we apply Wasserstein distance for measuring the distance of multi-dimensional distribution.Based on the passengers’ visit to metro stations,the metro stations are clustered according to the familiarity of passengers,which can provide valuable insights into sign design,streamline planning,and advertising inside stations.(2)By investigating the spatial and temporal correlation of individual travel,we identify group travel.Three indices are proposed: the number of overlapped traces,spatial diversity,and passenger user adhesiveness.According to the statistical characteristics of passenger travel,methods to find the group passengers thresholds are established.Based on the thresholds,whether or not two passengers are group passengers are determined.Afterward,whether trips are with companions and the number of companions for each trip can be obtained.Then we study the spatial and temporal characteristics of group trips.Combined with house price,the interaction between behavioral patterns and socioeconomic attributes of group travel is investigated.Using tensor decomposition technology,the passenger flow is divided into several types according to the origin,destination,with or without a companion,departure time,and travel date.The results show that in the time dimension,passengers tend to travel together in the afternoon and on weekends;from the spatial perspective,areas with moderate housing prices are the smallest source of group travel,while areas with relatively low housing prices are the most popular.In addition,based on the group travel behavior,a social network is constructed.After investigating its static structure and comparing it with online social networks,we conclude that,unlike online social networks,the social connections formed by group passengers were independent of network size.(3)Finally,based on the constructed individual contact network and passenger space trajectory information,the dynamic process of the individual and population coupled network is studied.Based on the constructed individual contact network,combined with the location information,a two-layer coupled network that can fully utilize multiple data sources is established.A disease transmission model considering both mobility and contact characteristics is proposed.Through the case study in Beijing,multiple scenarios were constructed based on control measures to verify the accelerating effect of individual contact networks on disease transmission.