Research On Analysis And Prediction For Traffic Data Of Wireless Networks

With the continuous development of wireless communication technologies,the number of users and access devices in wireless networks has been increasing constantly.As a result,wireless networks have become more complicated and energy consumption also has increased.By analyzing and forecasting wireless network traffic data,the trend of future wireless network traffic data can be captured and the service providers can adjust the system configuration and change the system parameters to meet the future traffic service requirements,at the same time they can make rational use of the existing network resources,maintain the network and optimize the existing network,so as to keep the wireless network in a good running state.Using the prediction results of wireless network traffic data to support the design of the sleeping of the base station can effectively reduce the network energy consumption.Therefore,the analysis and prediction of wireless network traffic data and the use of data for base station sleeping have been widely studied in academia.In this paper,based on the actual traffic data of wireless network,we study the time series forecasting method and propose an energy saving scheme based on the forecasting results.First,this paper introduces the background and significance of the study,as well as the traditional time series prediction method and the emerging echo state network model,and the standard modeling steps of them.Then,the paper analyzes the data in the real network,marking the data into strong pattern data and weak pattern data according to its periodicity and randomness,and makes use of the Seasonal Auto Regressive Integrated Moving Average Model to predict.Afterwards,the periods of the sequence were obtained by spectral analysis,and the new time series were constructed by using the periods and time as the exogenous variables.The single-step prediction result shows that prediction using ESN with exogenous variables is better than SARIMA.Furthermore,this paper considers the use of prediction results to study the base station sleeping.For the base station sleeping we need to predict the next 24 hours' base station traffic data prediction,so we need to consider the multi-step prediction model.Based on the multi-step traffic prediction results,a base station sleeping mechanism that guarantees base station coverage is proposed,which enables some base stations sleep when the load is low and the surrounding base stations to bear the traffic load of the sleeping base stations so as to save energy.Simulation results show that the base station sleeping mechanism can effectively reduce network energy consumption.