Traffic Flow Prediction Of Road Network Moving Objects Based On Sports Pattern Mining

With the construction of the intelligent city rapidly developing,the traffic flow has become the focus of the urban monitoring and management.The traditional monitoring the traffic flow is generally obtained by the sensors on the road,but the sensors can not cover all the road networks,and the maintenance costs are high.So how to realize monitoring and forecasting the traffic flow by using the continuous vehicle trajectory data has become the focus of attention.The track data is not only huge in number,and is required to be processed in the real time in order to make the accurate real-time decision in the rapidly changing traffic environment,which brings a new challenge to how to make full use of the vehicle track data on the road network to accurately predict the traffic flow on the road.In addition,the characteristics of space and time are not fully excavated by the method of the traffic flow prediction based on the trajectory data,which leads to the low prediction accuracy.In order to improve the timeliness and the accuracy of monitoring and forecasting the traffic flow,this paper presents a medtod of the traffic flow prediction based on the track data of the moving vehicle.Firstly,the method uses the distributed hidden Markov model to match the track data of the moving vehicle to the map.Secondly,the traffic flow sequence was calculated according to the road number and the time period.Finally,a parallel long-time neural network(Long Short Term,LSTM)is used to mine the short-time and periodic features of the data,and the features are used to predict the traffic flow.The main contributions of this paper are as follows.1: Considering the amount of the vehicle track data,the traditional methods for the datum analysis can not realize the effective analysis of massive data,so in this paper,a map matching algorithm of hidden Markov model based on the hadoop platform is proposed,which can deal with the map matching the problem of massive spatiot emporal data more quickly and accurately.2.In this paper,the parallel LSTM neural network is used to predict the traffic flow of the road,and the short-time and periodic characteristics of the road are excavated,and the optimal parameters are determined in the process of model training,which greatly improves the accuracy of the prediction.3.The prototype system of the road traffic flow prediction is designed and compared with other traffic flow forecasting models,which is proved that the proposed method is feasible and effective.