Research On The Prediction Algorithm Of Bus Arrival Time

With the development of traffic,road congestion,vehicle exhaust emissions caused by environmental pollution and other issues are becoming more and more serious.Compared with private cars,public transport has many advantages,such as large capacity,low energy consumption and relatively small exhaust emissions,which is of great significance to alleviate the above-mentioned problems.Therefore,to improve the passenger satisfaction of bus is the necessary way to attract passengers.The main content of this paper is the prediction of the time interval from the starting station to the terminal station of the bus.At present,bus companies in our country mainly rely on experienced staff to estimate the return time of vehicles and then carry out vehicle scheduling.However,due to the lack of auxiliary prediction algorithm,only empirical estimation often results in large errors and erroneous scheduling decisions.The existing bus arrival time prediction is usually used to predict the number or time of the nearest bus arriving near the station,which belongs to the inter-station forecast.If the accumulative method is directly used to predict the terminal,error accumulation may occur.Therefore,it is necessary to study a method which can predict the time interval from the starting station to the terminal station,both in scientific research and in the engineering sense.Firstly,the paper introduces the theory and application research status of vehicle arrival prediction technology at home and abroad,and introduces the principles,advantages and disadvantages of several common bus arrival prediction techniques.Secondly,this paper introduces the data processing and analysis process of influencing factors.Finally,two kinds of prediction algorithms are proposed:R-GBDT and CTA.R-GBDT uses feature selection component and model parameter adjustment component to provide feature combination and parameters in accordance with line characteristics for prediction component The fusion component performs fusion prediction on the results of other components for the final time interval.CTA uses spatio-temporal component to capture the temporal dependence and spatial correlation of things with the help of the ConvLSTM algorithm.Using attribute components to deal with the external attributes of things by embedding and standardizing methods.The output of the spatio-temporal component and the attribute component is fused by the fusion component to predict the total travel interval.