Research And Implementation Of Traffic Model Based On Wireless Network Signaling

With the rapid growth of the number of mobile terminals,a large amount of wireless network signaling data is generated in the cellular network.From the massive data,the commonality and regularity of user travel are analyzed,which can provide technical guidance and Basis for decision.The assumptions of traditional traffic models are too ideal,and most of them do not use measured data for modeling and verification.Therefore,they have insufficient versatility and cannot be directly extended in the existing traffic network system.In response to the above problems,this paper uses the measured data of wireless network signaling to conduct traffic model research,and the main contributions are as follows:First,based on the design concepts of long-short-term memory network model and gated recurrent network model,a group normalized grid recurrent network traffic flow prediction model is proposed.The proposed model improves the scale of the model's data source and the speed of prediction convergence through group normalization when data is input.At the same time,the calculation process of the internal circulation unit is simplified,and each iteration does not need to wait for the output of the previous hidden layer,and stores and updates the required data through the memory unit.In addition,for the noise data existing in the original data,the data set is gridded,so that the model screens the source data set to avoid the adverse effects caused by the noise data.The simulation results show that the group-normalized grid recurrent network model proposed in this paper has good prediction performance,and at the same time can reduce the prediction calculation workload of the traditional model and improve the prediction processing efficiency.Secondly,this paper adopts the training-feature-retraining modeling idea to propose a traffic travel mode recognition model based on convolutional neural network.Considering that the wireless network signaling data set has label-free characteristics,the model is first trained using the open source GPS data set to obtain model recognition features.Then,the model recognition features are put into the deep neural network of the wireless network signaling data set to continue training,so as to obtain user travel classification results of different categories.Simulation results show that the model has good user classification performance,and solves the problem of low recognition rate of traditional models.Finally,the performance of the proposed traffic model and traditional model are evaluated and analyzed.For traffic flow prediction scenarios,the performance comparison between the proposed traffic flow prediction model and the traditional model for traffic flow prediction is made.Experimental results show that the traffic flow prediction model proposed in this paper has better prediction accuracy and better model performance.Aiming at the traffic travel user identification scenario,the travel mode identification model submitted in this paper is compared with the traditional model.The experimental results show that the user classification accuracy of the proposed model is higher and the model recognition performance is better.