Deep Learning Based Traffic Flow Prediction For Wireless Networks

With the rapid development of mobile communication technology,the mobile Internet has been applied to many fields,such as payment,catering,transportation,multimedia and so on.The traffic flow data of wireless communication presents explosive growth,but the network resources are limited.How to meet the above service requirements under limited resources has brought great challenges to the mobile communication network and network operators.Through the prediction of wireless network traffic flow,it helps to reasonably and effectively allocate network resources,such as spectrum resources and channel resources,to provide users with more stable and efficient services and ensure network performance.Wireless network traffic flow prediction methods are varied.Among them,neural network is widely used in this field.However,the traditional neural network has limited utilization of time correlation of time series data,so it can not meet the accuracy requirement of current resource allocation.Therefore,this paper mainly studies how to use deep learning to improve the accuracy of wireless network traffic flow prediction,so as to make full use of wireless network resources.In this paper,two methods of wireless network traffic flow prediction based on deep learning are proposed.Firstly,the method of wireless network traffic prediction based on stacked autoencoder network(SAE),which mainly utilizes the feature extraction ability of stacked autoencoder network to effectively predict traffic flow.Secondly,the method of wireless network traffic flow prediction based on long short-term memory network(LSTM),this method uses the long short-term memory network to predict the wireless network traffic flow,making full use of the time correlation of the wireless networks traffic flow data.It has a relatively long memory function for the input data,and can store the incoming data in the memory cell.When predicting traffic flow,long short-term memory network can predict future traffic flow based on the data in the memory cell and the input of the current time.Simulation results show that compared with the results of wireless networks traffic flow prediction based on stacked autoencoder network,the prediction accuracy of wireless networks traffic flow based on long short-term memory network is higher.