Research On The Identification And Prediction Method Of Urban Public Transport Transfer Behavior

The transfer between metro and bus is a core component of the urban public transportation system,which increases the accessibility of metro and bus networks and provides reliable,fast,and economical services for passengers.It is important to accurately identify the transfer behavior of buses and metro,explore the characteristics of transfer volume,transfer rate,and transfer time at each station,investigate the spatial and temporal effects of various factors on transfer characteristics,predict the information of transfer characteristics at each station,and ultimately improve the efficiency and service level of the transfer system,increase the attractiveness of public transportation,and thus alleviate the traffic congestion problem prevailing in major cities.It is of great significance to provide guidance.However,existing transfer identification methods are difficult to accurately identify the transfer behavior of bus-to-metro mode and have difficulties in mining the transfer characteristics of each station.Meanwhile,it is still unknown how each factor affects the transfer features in time and space.Existing studies do not propose accurate prediction methods for transfer features,and cannot provide transfer passengers and transportation practitioners with transfer prediction information at each station.Therefore,it is urgent to propose a twoway transfer identification method for public transportation systems,to explore the transfer characteristics of each station,to analyze the spatio-temporal influences of the transfer characteristics,and to accurately predict the transfer characteristics of different stations.Based on the transfer feature mining and analysis of urban public transportation systems,this paper conducts research on transfer identification,feature mining,and application of different transfer modes,analyzes the transfer feature distribution characteristics of metro-tobus and bus-to-metro modes,determines the differences in the spatio-temporal influence of each factor on transfer features,summarizes the existing problems,proposes countermeasures,improves the accuracy of transfer identification methods and It provides a reference for the analysis of the spatio-temporal influence factors of transfer features,a feasible method for the accurate prediction of transfer features,and accurate transfer prediction information for transfer passengers and transportation practitioners,which in turn provides effective support for the optimization of public transportation transfer systems in smart cities,public transportation scheduling and subsequent management control.The main work and research results of this paper include the following aspects.(1)The traditional transfer recognition method has difficulties in identifying the transfer behavior of bus-to-metro mode and has not been validated by field data.In this paper,we propose a transfer recognition method based on the transfer time threshold and transfer space threshold for the metro-to-bus mode and bus-to-metro mode,which has been validated by field survey data.The method can identify the transfer behavior of different transfer modes at each station within each hour from multi-source data such as large-scale smart card data and bus operation trajectory data,and complement the travel chain of transfer passengers to accurately mine the transfer characteristics of different transfer modes at each station within each hour.Compared with other methods,the transfer data identified by the transfer identification method proposed in this paper has a better fit with the survey data and higher identification accuracy,which can support the research of public transportation transfer feature influence factor analysis and prediction methods.(2)The existing studies did not analyze the spatio-temporal relationship between transfer characteristics and influencing factors,nor did they explore the advance and lag effects of each factor on transfer characteristics.Regression models are used to investigate the effects of various factors on transfer characteristics during weekdays,weekends,holidays,and typhoons for different transfer patterns,and the regression performance of different methods is compared.The results show that the generalized Poisson regression model has a better performance compared to other regression models;weather factors have a significant effect on the number of transfers with one hour advance or three hours lag.These findings provide data support for public transportation metro station improvement and transfer service level improvement.(3)The existing studies do not analyze the spatial heterogeneity of transfer characteristics,and the spatial and temporal effects of various factors on transfer characteristics are unknown.Multiple spatial regression models are used to explore the spatial relationships between transfer characteristics and influencing factors for different transfer modes,to explore the spatial heterogeneity of transfer characteristics,and to explain the spatial and temporal influences of each factor on transfer characteristics.The results show that the geographically weighted regression model has the best performance compared with other spatial models.Weather factors,built environment,socio-economics,POI data,and business activity intensity have significant spatial effects on transfer characteristics and there is spatial heterogeneity.To address the negative effects of each factor on the transfer characteristics,this paper proposes improvements and countermeasures to support the management and planning of public transportation transfer systems.(4)The current study does not propose an effective prediction method for transfer time.A Bayesian low-rank matrix decomposition-based transfer time prediction model is proposed for the spatial and temporal distribution characteristics of transfer time.The results show that the prediction model proposed in this paper can accurately fill in the transfer characteristics data under different missing scenarios and different missing rates,accurately predict the transfer time data of each metro station,and the filling and prediction performance is better than the comparison models,which can provide accurate transfer time prediction information of each metro station for transfer passengers and transportation managers,and can support the monitoring and analysis of the transfer situation at each station of the public transportation system and Optimization of scheduling and other services.(5)The current study does not propose a suitable prediction method for transfer volume,nor does it consider the influence of multiple factors on transfer volume.A multivariate transfer prediction model based on machine learning is proposed by considering the effects of several factors on transfer volume.The results show that the proposed prediction model can accurately predict the transfer volume at each metro station based on accurately filling the transfer volume data under different missing scenarios with different missing rates.Compared with statistical models and other machine learning models,the proposed multivariate transfer prediction model has better performance and can accurately predict the transfers at each station,which lays the foundation for subsequent research on interpretable machine learning models in transfer feature prediction.In summary,based on the need for transfer identification and prediction under different scenarios,this paper proposes a two-way transfer identification method for urban public transportation systems,a spatio-temporal regression model for transfer feature influencing factors,and a prediction model for transfer features.These models and methods have their characteristics and complement each other,covering various aspects such as transfer identification,spatio-temporal influence factor analysis of transfer features,transfer feature data repair and prediction,etc.They are organically combined to form a set of high-precision algorithms applicable to urban public transportation systems,providing a complete solution for transfer feature mining,analysis,and prediction of urban public transportation systems.