Research On Traffic Flow Prediction Algorithm Based On Bidirectional Cyclic Neural Network

In recent years,with the continuous improvement of the social level,private cars have become an indispensable means of transportation for people,but the the convenience of the cars also caused a lot of pressure on the traffic,following traffic congestion and environmental pollution,Traffic flow prediction algorithm has been widely studied,it is hoped that drivers can make road planning ahead of time and reduce traffic pressure.Although there are many traffic flow prediction algorithms based on cyclic neural networks,most of them are based on past data to predict the future,without considering the future data to predict past data to achieve the goal of predicting the future.This thesis conducts in-depth research on this aspect.The main work of this paper is as follows:Firstly,the BiLSTM(Bi-directional Long Short-Term Memory)algorithm is applied to the field of traffic flow prediction,and the traffic flow data is multiplexed by two layers of LSTM.Considering that the traffic flow data discarded at any time has a certain influence on the subsequent data,the feature extraction is performed using the forward and reverse LSTM,so that the output value of each time step depends on the value of the two-layer algorithm,and a more accurate traffic flow prediction model is obtained,so as to improve the accuracy of the prediction results.Like the same principle,the Bi GRU(Bi-directional Gated Recurrent Unit)algorithm is applied to the field of traffic flow prediction.Secondly,in order to further improve the accuracy of traffic flow prediction and the realtime ability of dealing with sudden events,this paper proposes DBiLSTM(Double Bidirectional Long Short-Term Memory)algorithm for traffic flow prediction.DBiLSTM not only considers the impact of past data characteristics on the future,but also considers the impact of future data characteristics on the past.On the basis of BiLSTM,the two BiLSTM prediction models are trained and obtained at the same time,so that the prediction of the same day and the past can be improved by linear fusion.The same principle proposes DBi GRU(Double Bi-directional Gated Recurrent Unit)algorithm for traffic flow prediction.Finally,through the design experiment simulation,four algorithms are used to predict the traffic flow and the prediction performance is evaluated.The experimental results show that the BiLSTM algorithm improves the accuracy of low-peak prediction compared with LSTM in traffic flow prediction,and the speed also has improved;DBiLSTM improves processing time and processing accuracy in mutation event processing compared to BiLSTM and LSTM.Compared with GRU,Bi GRU algorithm improves the sensitivity of processing data in the prediction of low peak period,and the prediction accuracy is slightly improved.Compared with Bi GRU and GRU,DBi GRU does not improve much in prediction accuracy,but there has been a significant increase in the speed and schedule of dealing with emergencies.In general,the improved algorithm has the characteristics of fast processing speed,high prediction accuracy and strong processing capability of dealing with emergencies in traffic flow prediction.