Research On Bus Arrival Time Prediction Based On Vehicle Networking System

With the development of intelligent traffic, many scholars at home and abroad have turned their attention to the study of the arrival time prediction of public transport. And that providing a more accurate and reliable bus arrival prediction time is one of the most concerned traffic information for travelers.This paper establish a higher accuracy and reliability of the bus arrival time prediction model through analyzing and selecting of the bus arrival time of various static and dynamic factors which influences the bus arrival time. The content mainly includes several aspects as follows:Firstly, combine the process of data whose the vehicle networking system collects the driving state information data of public transport vehicles, so that we can get the details information the vehicles arrives in the station. This paper analyzes and selects the various important components which impact the bus arrival time to quantify those input variables.Secondly, based on selecting and getting the important components which influence the bus arrival time, the Deep Learning Model is brought and established to predict the public vehicle arrival time after the analyzing the shortcoming of the previous existing traditional model. The training steps and parameter setting of the prediction model are given.Finally, the thesis apply the Deep Learning Model to predict the vehicles arrival time in a selected route which comes from the Harbin bus company for verifying the feasibility of the forecast model. Otherwise the BP Neural Network Model and Support Vector Machine Model are established respectively to comparative analysis the forecast effect of the established model. We get the conclusion that the Deep Learning has a higher accuracy and a shorter training time than the other two model by analyzing the experiment.