Research On Regional Travel Destination Choice Model Based On Cellular Signaling Data

Regional traffic demand model can quantitatively guide the construction of regional traffic system and support the development of regional traffic integration.However,due to the difficulty of traditional survey methods in regional level,and the difficulty of obtaining travel information in residential areas,data bottleneck exists in model construction and parameter calibration.At the same time,regional travel has a long distance and a large span,and the travel distribution rules generated under this structure are quite different from those of urban travel.It is difficult to highlight the regional traffic characteristics by adopting a broad and general traffic distribution model.Therefore,it is necessary to combine regional travel characteristics to propose appropriate travel distribution model.In the travel distribution model,destination choice model belongs to the category of discrete choice model.It is constructed based on utility maximization theory.Its form is more flexible and extendable than gravity model,and can explain the influence of personal attributes,social economic and demographic characteristics on travel behavior.Therefore,based on the signal data of mobile phone,this paper extracts the regional travel characteristics of travelers and proposes a data input method as a destination choice model,breaking through the difficult problem of obtaining basic data relying on traditional methods.Based on large sample and complete travel distribution information,the destination choice model was improved to accurately describe the destination choice behavior of regional travelers and improve the forecast accuracy of destination choice model.Firstly,this paper carries out the research on regional travel feature extraction based on cellular signaling data.By defining the regional travel,this paper analyzes the applicability of cellular signaling data in the regional scale.For a large number of original cellular phone signaling data,preprocess invalid,abnormal and drift data.A set of regional travel feature extraction algorithm based on rules and DBSCAN clustering is constructed to identify travelers from regional travel stop point,start and end point,regional travel time and travel purpose.Then,based on the regional travel characteristics identified by mobile signaling data,a destination choice model is constructed.Before the model is built,the standard deviation of average travel distance and average travel time is selected as the index to measure the difference of regional travel characteristics.The travelers are divided into working travel and non working travel groups,and the destination choice models are constructed respectively.By determining the destination spatial choice set and choice limb scale,the utility function of the model is constructed based on the relevant influencing factors of destination choice behavior.Secondly,the destination choice model is improved based on large sample and complete trip distribution information of signal data identification of mobile phone.Considering that the common items in utility function are only related to the destination cell,and the choice of destination cell by traveler depends on his own departure cell,an estimation method for the specific constant items of departure cell and destination cell is proposed,which supports the estimation of specific constant items of different choice limbs based on a large number of regional travel samples.Finally,it makes an empirical study on the destination choice model.Based on the mobile signaling data of Chengdu and Deyang from December 5 to January 5,2021,this paper studies the travel between Deyang and Chengdu.From the perspective of average travel distance and travel time,the regional travel characteristics of working travel and non working travel groups are the most different.For work travel,the number of scenic spots in the community,the number of large shopping malls and leisure service facilities have no impact on the choice of destination,while the number of jobs,the number of industrial enterprises,population density and per capita GDP have a strong positive effect,and the work travel group is more sensitive to time impedance.Through model comparison,it is found that the prediction relative error of the improved destination choice model is about 6.8%,which is 18% higher than that before the improvement.The performance of travel time distribution and average travel time is significantly better than that of gravity model.