Research On The Optimization Strategy Of Mobile Fronthaul Network Based On Machine Learning

With the evolution of communication technology from 4G to 5G,mobile network scenarios are diversified,user traffic is increasing,and user mobility is strengthening.However,the traditional bandwidth allocation strategy is to statically allocate sufficient resources for the peak load of link and mainly schedule after the arrival of the service,which results in the low efficiency of network resource utilization and the lag of network scheduling.Therefore,the current mobile network urgently needs the traffic awareness and the pre-allocation adjustment of resources to realize the dynamic optimization of resources.As an important research field of artificial intelligence,machine learning provides a variety of prediction models.Therefore,how to realize the effective integration of machine learning and mobile network resource optimization is a problem that network operators need to face in new network scenarios.In this paper,a mobile fronthaul network resource optimization strategy is proposed based on machine learning,in which a traffic prediction network model is built based on machine learning.After that,fronthaul bandwidth is dynamically allocated based on the prediction results,and finally the optimization experiment is completed on the hardware platform.The contributions and innovations of this paper are as follows:(1)A traffic prediction network model is built based on LSTM network,and a mobile prequel network bandwidth dynamic allocation strategy is proposed based on traffic prediction results.This paper fully analyses the characteristic of the base station traffic data and builds three kinds of prediction model based on ARIMA algorithm of statistical model,the SVM and the LSTM algorithm respectively,training the prediction model by inputing the real historical traffic data and outputing traffic prediction in future.By comparing the optimal prediction model,the simulation results show that the prediction model based on the LSTM network shows the best optimal fitting effect in multiple cell scenarios.Based on the traffic prediction results,this paper proposes a mobile fronthaul network bandwidth dynamic allocation strategy,which realizes the bandwidth dynamic adjustment in advance based on traffic forecast.(2)The machine learning-assisted Converged Edge Access Network Platform testbed is designed and completed,and finally the dynamic optimization experiment of mobile fronthaul network resources is completed based on this testbed.This paper proposes to embed the artificial intelligence module based on TensorFlow plug-in into the testbed to provide intelligent decision-making,by comprehensively considering the great prospect of artificial intelligence technology in SDN control platform,and completes the corresponding design and development work.Based on the above design and development of the hardware testbed,the mobile fronthaul network resource optimization experiment based on machine learning is completed,and the intelligent traffic prediction algorithm and the fronthaul bandwidth dynamic allocation algorithm are integrated in the hardware testbed effectively.