Prediction Of Bus Service Time At Platform Based On Bus Big Data

With the development of national economy and the continuous promotion of the strategy of building a strong transportation country,the concept of green travel has been deeply rooted in the hearts of the people,and public transport travel has gradually been accepted by the general public.As the basic unit of public transport network,bus stops are widely distributed in the urban road network.The service time of bus stops seriously affects the operation time of the whole line.Accurate prediction of bus stop service time is conducive to the release of bus information and intelligent integrated scheduling.In this thesis,the service time of bus stops is studied.Based on the analysis of the factors that affect the service time of bus stops,the service time prediction model of bus stops is established.Firstly,the thesis classifies the IC card data,uses the single factor analysis of variance and logistic regression model to analyze the five influencing factors of user type,station type,travel time,holiday and weather type,and gives the prediction model of passenger boarding time based on Logistic regression.It is found that the type of users and the type of stations have a positive correlation with the boarding time,while the travel time and holidays have a negative correlation with the boarding time.The order of influence degree of each factor is: user type,site type,travel time,holiday.Secondly,the site is divided into two types: leisure area and work area,and nine influencing factors are constructed.The commercial area and school campus are selected as typical sites.Through the establishment of Poisson regression,NB regression,ZIP regression and ZINB return,the relationship between the number of waiting passengers at the site and the influencing factors is explored.The results show that ZINB model can best fit the data of bus waiting number.The factors that influence the business district are bus number,actual time distance,operation time and rainy day.The significant influencing factors of the campus are: the number of bus routes,the actual time interval,the running time,the weekend,the number of people in the car and the number of people in the first half of the car.Finally,the thesis establishes the LSTM-RNN network to predict the service time of bus stops,and establishes two different BP neural network models for comparison according to whether or not considering the line differences,and selects Xi’an bus stops as an example.The results show that the prediction error rate of LSTM-RNN is always the lowest in the four groups of examples,no matter in the whole or in each period,The MAE value of LSTM-RNN model is as low as 2.52,and the MARE value is as low as 23.05%.