Research On Network Traffic Prediction And Application Based On Machine Learning

With the development of computer information technology,the functional scenarios and architecture of the network are becoming more and more complex,and the network traffic is exploding,showing many new features,which pose new challenges to the performance improvement and stability maintenance of the network.In order to enhance the network operation rate and improve the utilization of network resources,this paper combines the machine learning algorithm to classify and predict the traffic based on the traffic multi-scenario modeling,and builds the SDN simulation platform for algorithm application,which has theoretical and practical significance.Firstly,the four typical traffic scenarios of backbone network,data center,edge network and large emergencies are combined with previous research work to establish Poisson,MMPP,self-similar and Pareto mathematical models.The characteristics of traffic are studied in depth and the reference size of the route buffer is taken out.These results are double verified by real traffic analysis and SDN simulation.Secondly,on the basis of traffic modeling,the lightBGM multi-classification algorithm is used to classify Poisson,MMPP and self-similar traffic data,and the data is pre-processed to extract features.The parameters are continuously optimized during training to achieve better classification results.Next,the data processing and prediction research is carried out for the self-similar traffic with strong bursts.The LSTM algorithm which can capture the long-distance dependence in the sequence is used for prediction,and the parameters are selected to achieve better prediction results.Finally,the ONOS controller is used to build the SDN platform,and the function modules of information collection,traffic classification,traffic prediction and route adjustment are established respectively.The whole process from information collection to traffic classification prediction and network optimization is completed,and the algorithm application is explored and practiced.