Study On The External Travel Characteristics Of Different Populations Based On Phone Signaling Data

The analysis of the characteristics of people’s external travel is one of the key technologies in the planning,operation and management of urban external transportation system.In view of the shortcomings of the existing research on the recognition and algorithm of specific external travel behavior,this paper takes the external traffic network of Kunshan as the research object,based on the actual collected signaling data of Kunshan mobile phone,and adopts machine learning technology,starting from three main links of external travel behavior recognition,external travel mode discrimination and crowd type identification Finally,this paper makes a deep and effective analysis on the characteristics of different groups of people.This paper is mainly from the following three aspects:Firstly,based on the signaling data of mobile phone,this paper studies the external travel identification method of mobile phone users.Through the analysis of mobile phone users’ external behavior mode,the external behavior can be divided into five types: inbound,outbound,transit,inbound round-trip and outbound round-trip.Combined with the temporal and spatial distribution characteristics of mobile signaling data when users travel abroad,the problem of external travel behavior recognition and external travel mode discrimination is respectively transformed into the problem of state recognition and map matching under the time sequence of mobile signaling data.Based on this,an algorithm of external travel behavior recognition based on sliding window is proposed,which provides the related statistical characteristics of external travel for the following crowd classification and recognition Sign.Secondly,this paper classifies and identifies the populations based on the improved kmeans clustering algorithm.On the basis of dividing the population into three categories:resident population,floating population and visitors,combined with the external travel frequency of mobile phone users and the distribution of duty and residence,the population is classified in detail.For the extracted candidate feature set of crowd type recognition,the unsupervised distance entropy method is used to select the feature set,and the optimal feature set is composed of seven fields,namely,the total number of days appearing,the number of days appearing in the residential period,the number of days appearing in the working period and the number of days traveling outside.On this basis,the improved k-means clustering algorithm is used to spatiotemporal cluster the feature matrix composed of the best feature subset,and the clustering effect is best when the number of clustering centers K is set to 7according to the DBI coefficient.The clustering results show that the population can be roughly divided into seven categories,such as transit passage population,short-term visit population,medium and long-term business trip population,and so on.Finally,this paper compares and analyzes the characteristics of the identified groups.This paper analyzes the basic attributes of all kinds of people from the aspects of mobile phone region,live and work place distribution.Combined with the time type of date,it analyzes the external travel traffic volume and external travel traffic mode selection of all kinds of people from three levels of working day,non working day and holiday.The results show that there is a close communication between Kunshan and Shanghai,mainly in Huaqiao area,and there is an obvious imbalance between live and work in this area,with a large amount of cross-border commuting traffic.In addition,visitors mainly enter and leave Kunshan through Kunshan south station,while cross-border commuters prefer to choose subway and road travel.