Research On Bus Arrival Time Prediction With GPS Data

In order to promote the sustainable development of traffic, it is necessary to give priority to the public transport. Bus arrival time prediction service can improve the level of public transport services. Therefore, we do the research on bus arrival time prediction technology with the GPS data.Since the running bus has been affected by many environmental factors, realizing accurate prediction of bus arrival time is a complex and difficult problem. In order to solve the problem many researchers have put forward a lot of model on bus arrival time prediction. The main existing problems of these models include:without considering the real-time traffic condition, model design based on experience of designer, over fitting problem, easy to fall into local optimum, can not meet the actual situation because of based on certain assumptions, can not meet the real-time application requirements because of high time complexity, too less factors considered, error accumulation because of unreasonable model structure.After doing a survey on bus arrival prediction model, we chooses support vector machine as the theoretical basis of prediction model which will be put forward in this paper. As a new machine learning algorithm, support vector machine can model complex nonlinear problem, and model design does not depends on the experience and any assumptions, meanwhile, there not exist problems about over fitting and falling into local optimal. When doing prediction with the support vector machine model trained, the time complexity is low, which can satisfy the requirement of real-time application. The original training set is too large to train the model of support vector machine, so this original training set will be divided into many small training set, and these small training set will be structured by a multi-way tree, which help to avoid the error accumulation problem. What’s more, in order to improve the prediction accuracy, several factors are considered, including the current traffic condition.Firstly,in the paper, some analysis on bus running time is made, and a scheme of bus line segment processing is put forward; secondly, with consideration on the holidays, weather, road, the peak time and other factors, a simple bus arrival time prediction model and a relatively complex bus arrival time prediction model based on ε-SVR are put forward; finally, with the real GPS data, the prediction accuracy of the model proposed are evaluated, and the results showed that the prediction accuracy is high; In addition, serveral key algorithms of data preprocessing are also put forward: the algorithm to determine the boundary of the bus station, the algorithm to determine the time of entering and leaving the station, the algorithm to determine the departure time of the bus. On the background of Shenzhen bus, based on the research content of this paper a software system is developed, which can provide the bus arrival time prediction of service, and the some application value is created.