Mining And Analysis Of Public Transportation Commute Travel Characteristics Based On Multi-source Data

With the increasing urbanization process and the rapid increase in the number of motor vehicles,the congestion of peak roads in the morning and evening has become the primary ill disease that plagues the transportation of large,medium and small cities in China,which has seriously lowered the happiness index of residents.As the main force of the city's morning and evening peak travel,the commuter crowd deeply explores and analyzes the travel characteristics and time and space characteristics of the commuters,and perceives the characteristics of commuter travel environment.It is of great significance to improve the service level of public transport,induce passenger flow and ease peak congestion.At present,the study of public transportation and travel has made it impossible to grasp the real commuter travel characteristics and passenger flow patterns by means of questionnaire data,based on the commuting research of credit card data,the commuter identification method is simple,the extraction result is not accurate,and the feature analysis is not deep.In this paper,based on the multi-source bus data as the research data foundation,two novel,efficient and accurate commuter identification methods are proposed,which deeply analyzes the commuter travel characteristics and travel environment.The specific research contents are as follows:(1)Multi-source bus data processing level,corresponding solutions are given for the lack of GPS station data and the inconsistency between the IC card swipe system and the GPS system clock.Then,using data fusion technology,the specific steps and methods of vehicle site matching,getting off site inference,and transfer site identification are given,and the OTD data of passenger bus travel trajectory is restored,which lays a foundation for later commuting identification and feature analysis.(2)Commuter recognition model based on machine learning algorithm,Firstly,feature engineering,selecting the best identification features and giving each attribute eigenvalue processing method,Then,based on this,the questionnaire is developed to obtain training data,and the best machine learning algorithm is selected.The improved cost sensitive GBDT(gradient lifting decision tree)algorithm is used to train the classification model,Finally,the corresponding features are extracted from the real credit card data,and the model is substituted into the model to complete the commute identification.(3)Commuting travel extraction based on closed bus travel chain,Firstly,the definition of closed public transportation travel chain is given,and the OTD-based travel topology map is constructed,Then,the depth-first closed loop search and the similarity association are performed,and the closed bus travel chain is extracted from the incomplete and fragmented travel track information,and starting from the commute travel time and space law,develop commuter travel chain screening rules to complete commuter extraction.Finally,a comprehensive comparative analysis of the advantages and disadvantages of the two commuting extraction results and methods.(4)Research on the characteristics of public transportation,Firstly,based on the extracted public transportation results,the travel characteristics are analyzed from the perspective of the travel chain,and analyze the travel time and space characteristics,Then,according to the commuting time of the main considerations of travel,the commuting time impact model is established,and the influence degree of each factor is analyzed,and the specific improvement opinions are given.The paper is based on Zhuhai source bus data,using tools such as Python,Mysql database,ArcGIS,Spss,etc,the accurate extraction of public transportation and the multi-level analysis of features have been completed from the real credit card data,which provides a new idea for the study of public transportation.