Research On Mobile Network Traffic Prediction Based On Big Data

With the rapid progress and development of 5G technology,mobile communication technology has been widely applied to the Internet of Things,multimedia live broadcasting and other fields,entering the era of the Internet of Everything.The data flow of mobile communication service also shows a geometric growth,but how to meet the above mobile network traffic growth without causing the demand of network congestion,to China's mobile communication network and the three operators have brought advanced opportunities and challenges.By predicting the mobile network traffic,it is helpful to make reasonable and effective use of the mobile communication network resources,provide users with more stable and efficient services,and ensure the performance of the mobile communication network.Now forecast model method is varied,including Lassoregression algorithm model,neural network model and XGBoostalgorithm model has been widely applied in the prediction areas,this paper first adopts the three model to forecast the mobile traffic analysis,however,a single prediction model is relative to the robustness and accuracy of traffic prediction is limited,can't fully meet the requirements for the network traffic prediction accuracy and soundness.Therefore,this paper proposes a dynamic weight optimal weighting method combined with a single predictive mobile network traffic prediction method.In the single prediction model,Lassoregression algorithm model has the advantage of compression variables and stronger explanability.BP neural network prediction is suitable for nonlinear prediction of big data with higher prediction accuracy.XGBoosthas the continuous and non-linear statistical properties of handling data.The dynamic weight optimal weighting method uses the prediction model to estimate the average difference in real time,and then dynamically distributes the weights according to the optimal weight allocation principle.In this way,the prediction advantages of a single prediction model can be combined,and the traffic prediction can be effectively carried out,and the prediction accuracy and robustness can be improvedThe simulation results show that the MAE of the model prediction after the dynamic weight optimal weighting method is less than the three single models,and the minimum is 0.024.RMSE were all smaller than the three single models,and the minimum minimum was 0.02.R2 was all larger than the three single models,and the minimum maximum was 0.074.The data indicated that the accuracy of the fusion model was higher than that of the three single models.The standard deviations of the total accuracy of the fusion model were all less than those of the three models,which indicated that the robustness of the fusion model was higher than that of the three single models.