Bus Dwell Time Analysis And Prediction Methods

As an important node of public transportation network,bus stations is also the direct connection between suppliers and users of public services.A deep understanding of the pattern of bus dwell time not only benefits for understanding the bus operation rule more accurately,but also provides theoretical support for bus stops optimization settings,real-time bus scheduling and public transit signal priority strategy,which help improve efficiency.The paper summarizes recent researches about bus dwell time features and prediction methods both in domestic and overseas.Based on that,the paper puts forward an advanced multi-angle analysis method by using large number of bus dwell time data,which is collected by GPS and AVL technologies.This kind of method,which applying the mathematical statistics and data mining methods,mainly analyzes three aspects of bus dwell time in detail,namely dwell time distribution function,single site dwell time features of time and lines,multi-site dwell time interrelationship.Moreover,on this basis,the paper modifies ARIMA time series model and puts forward the real-time prediction method based on the data of the vehicle before.Accessing it for its accuracy,robustness,ultimately through the example analysis,the conclusion is that tihe adapted model has a more accurate result than the former ARIMAmodel;and even more accurate when the extreme values are eliminated.The model has a good performance in short time,which will bring a bright application prospect to the bus priority signal control system.