Research On Bus Arrival Time Prediction Based On LSTM-RNN

With the rapidly development of information technology,city administrators proceed with the construction of smart city.As an important component of smart city,intelligent public transportation system becomes more and more attractive.Bus arrival prediction is one of key technology to realize intelligent public transportation system.Corresponding prediction performance greatly influences user's travel experience.This paper study on the bus arrival time prediction method,while both GPS position calibration method and arrival time prediction model are considered.The specific contents are as follows:First of all,corresponding related works are analyzed.Then key points,which influence prediction accuracy,are discussed.After that,bus data set that is employed in this thesis is introduced,while a data preprocessing procedure is designed to remove invalid information.The impacts of bus cruising state are discussed.Then,based on bus cruising data set and driving route,a GPS position information calibration method is proposed.The GPS locating error and the reasons of causing the locating error are discussed,while typical scenarios are selected to develop corresponding calibration strategies.A Gaode map API platform is employed to verify the effectiveness of proposed method.The experiment results show that proposed method performs a high accuracy.Finally,an arrival time prediction method,which considers both real-time traffic state and history traffic state,is proposed.A LSTM-RNN learning model is constructed to realize bus arrival time prediction.The calibrated data set is used as input of learning model to test the performance of proposed method.Experiment results verify that proposed bus arrival time prediction method is valid.