Design And Realization Of The Method Of Forecasting Wireless Network Traffic In Base Station Cell

With the growing popularity of 5G networks and mobile devices,the demand for mobile data traffic is increasing.For network operators,it is significant to provide unhindered and ubiquitous high-quality services.Among them,it will be a challenge to build an accurate long-period prediction model of base station cell cell network traffic to guide operators to expand base station cell cell wireless devices.However,the burstiness and uncertainty of base station cell network traffic,so the base station cell cell network traffic has non-linear and non-stationary characteristics,which is a challenge for the long-term prediction of network traffic.Also base station cell traffic prediction shows some flexibility.Previously,base station cell traffic was predicted based on a single traffic sequence,while in base station cell network traffic,each cell corresponds to a network traffic sequence,so base station cell network traffic is composed with more than one traffic sequence.Based on the above characteristics,this makes the traditional statistical-based traditional model no longer applicable to the joint prediction and characterization of multiple base stations.The main work and contributions of this paper are as follows:1.The main methods of network traffic prediction are reviewed,and the advantages and shortcomings are analyzed,the characteristics of network traffic and related theoretical knowledge are introduced,and a long-term prediction method based on Prophet model is proposed for individual traffic sequence prediction.Empirical Mode Decomposition(EMD)and Ensemble Empirical Mode Decomposition(EEMD)are used to decompose the base station network traffic,and then the Prophet model is used to predict the decomposed The decomposed model is then predicted using Prophet model to solve the impact of non-smoothness on network traffic prediction and improve the model prediction performance.2.In order to adapt to the joint prediction of network traffic of multiple cells at base stations,a DeepAR-based deep learning model is proposed to predict the cell traffic of base stations,which can learn the features exhibited by different cells and update the neural network parameters.And based on this,we propose a manual feature calculation method based on Local Moving Average(LMA).Based on the a priori knowledge,the features are manually selected for the model.Compared with the traditional statistical models Prophet and EMD and EEMD combined with Prophet model,this model greatly improves the overall base station cell long-term prediction accuracy3.This study uses Prophet as well as EMD/EEMD-Prophet and DeepAR as traditional statistical models and deep learning models,respectively.The former has the characteristics of high interpretability and easy modeling,and can be quickly analyzed and predicted in single cell traffic sequence prediction.The latter has some advantages in joint prediction of multiple cells.Based on the above models,we design and implement the base station cell traffic prediction system.The system is developed and designed by Flask Web technology,which is provided to the network operation and maintenance personnel of base stations.The main functions of the system include data processing,traffic prediction,and history query.