Arrival Time Prediction Of Public Transport Based On Recurrent Neural Network

As an important part of transportation system,how to improve the intelligence of urban public transportation is very important to solve the problem of trafic congestion.In the process of building intelligent public transportation system,it is very important to improve the utilization rate of public transportation network resources and accurately predict the arrival time of public transportation vehicles.Accurate prediction of bus arrival time is a difficult problem in the field of public transport.Traditional forecasting methods mainly use arrival time and distance between stations,but not make full use of dynamic factors such as the number of passengers,dwell time and bus driving efficiency,which has a significant impact on bus arrival time.At the same time,in recent years,the combination of artificial intelligence new hardware and public transport is a new development trend to improve urban traffic intelligence.However,how to optimize the performance of the algorithm and how to implement the algorithm through a highly integrated platform(such as FPGA,embedded,etc.)also need to be discussed and solved.In order to overcome the shortcomings of traditional methods,this paper proposes a method to evaluate the efficiency of public transportation and a new method to predict the arrival time of public transportation based on the cyclic neural network(RNN)with two-stage attention mechanism.First of all,through the in-depth study of the model and algorithm of the arrival time prediction of the public transport vehicles,it is found that the performance of da-mn(recurrent neural network based on two-stage attention)is better than other classical prediction models.We choose to introduce attention mechanism to adaptively select the most relevant factors from heterogeneous information,and establish a prediction network based on da-rnn(recurrent neural network based on two-stage attention).Secondly,it is found that the accuracy of bus arrival time prediction can be improved by inputting dynamic factors into the arrival time prediction model.Therefore,this paper uses graph theory and the basic theory of complex network to model the urban bus stop network,and takes the network edge as the driving route between stops to analyze the bus driving efficiency,and puts forward a calculation method of the bus driving efficiency,as a dynamic factor.At the same time,a data set including dynamic factors is established.The experimental results show that the performance of various prediction algorithms(such as SVM,Kalman filter,MLP and RNN)is significantly improved after using dynamic factors.Finally,the RNN model with attention mechanism not only makes full use of historical information,but also captures the key features of prediction from static and dynamic factors.In this paper,the real data of Jinan bus driving is used for the experiment,and a variety of other methods under the same data set are used for the prediction experiment.Through the comparative analysis of the experimental results,the method proposed in this paper is superior in the data set provided by Jinan bus company.