Research On Departure Time Choice Behaviors Based On Travel Time Reliability

With the rapid development of China’s social economy and the accelerating pace of modern life,travelers pay more attention to the value of travel time,and the travel time reliability of urban rail transit has increasingly become an important indicator for passengers to measure travel efficiency and for traffic managers to evaluate operational efficiency.Passengers usually adjust departure time according to the travel time reliability of route and time period,therefore,it’s necessary to establish a systematic travel time reliability indicator system and analyze the influence of travel time reliability on passengers’ departure time choice.Under the network operation management of urban rail transit and based on AFC data,this paper studies the influence of travel time reliability on passengers’ departure time choice.The main work in this paper can be summarized as follows:(1)Based on the observed fluctuation of passenger travel time in Beijing subway,the travel time distribution and the travel time reliability are studied.The characteristics and preprocessing steps of AFC data are described,and the composition of urban rail transit travel time is analyzed.Based on the fluctuation characteristics of passengers’ travel time in Beijing subway and the existing research,five probability distribution models are selected to fit the travel time distribution of Beijing subway,and the Kolmogorov-Smirnov test method determine the optimal travel time distribution is Burr distribution.Then,the definition and significance of travel time reliability are explained and according to the pattern of Burr distribution,typical travel time reliability evaluation indicators are summarized from the perspective of operation managers and passengers,which formed a set of travel time reliability evaluation system.(2)Considering the disadvantage of calculating reliability by distribution fitting,a travel time reliability measurement based on predictability is proposed.The advantages of the travel time reliability measurement method based on predictability are analyzed.Then,the rationale of the Lempel-Ziv algorithm is described and the calculation steps and relevant formulas of the predictability method are given,including travel time series coarse graining,sequence compression,and predictability calculation.A case study of Beijing subway is carried out to study the travel time predictability of different types of OD,including the predictability results of the current period,the predictability results considering the past period,the predictability results of different feature days and the total predictability of OD.Finally,the LSTM is used to verify the feasibility of predictability method,and the results are compared with the reliability indicators results based on the Burr distribution.According to the Pearson correlation coefficient,there is a strong correlation between the predictability method and the reliability indicators,and it’s proved that the reliability measurement based on predictability can reveal the characteristics of time series,is not limited by the heterogeneity of indicators,and the calculation steps are efficient and simple.(3)Based on the Cumulative Prospect Theory,a model of commuter departure time choice considering the travel time reliability is constructed.Firstly,the main factors influencing the departure time choice are summarized and the process of departure time choice is analyzed.Then,the applicability of Cumulative Prospect Theory in departure time choice is discussed,and the passenger departure time choice model considering the travel time reliability is built based on the Cumulative Prospect Theory.Finally,a case study of model application for commuters in Beijing subway is conducted.The travel characteristics of commuters are descriptive analyzed,the Likelihood function and Genetic Algorithm are used to calibrate the parameters of the model.The results show that commuters’ risk preference decreases,they are more sensitive to loss,and their perception of objective probability improves when considering travel time reliability.