Research On Network Traffic Prediction Based On Time Series Model

With the rapid development of the Internet,the current network environment is becoming more and more complicated.This means that the network services and quality is prone to problems,the more the performance of the network will be affected.In the context of the triple play of radio and television networks,China’s radio and television network companies need to integrate and transform the existing network resources in order to enhance the market competitiveness.Therefore,to build a network management system for the operation and management of network equipment,System is currently facing the urgent need to solve the development of radio and television industry.Internet / SNMP network management system is currently the most widely used network management system framework,adaptability and scalability.Radio and television network is huge,complex structure,network fault maintenance and health management to enhance the network’s operational capacity is essential.Therefore,this paper will analyze and forecast the network traffic data of Hunan radio and television network using time series model in view of the complicated operation and maintenance of radio and television networks,the difficulty of resource scheduling and the monitoring of resource consumption.The traditional models used to predict the flow data are Markov process,Poisson process,AR model,MA model,ARMA model,ARIMA model,etc.These models are used to deal with the short correlation time series has a good effect,but can not have Long memory characteristics of the network traffic data for effective fitting and prediction.Therefore,it is important to find and study more suitable models.The FARIMA model is composed of the Hurst index and the ARIMA model.It is an extension of the ARIMA model.It has long correlation features and can use the parameter Hurst index to deal with the long memory part of the network traffic data.Therefore,the FARIMA model can be considered as the network traffic data Of the research model.Although the model selection is good,but in the process of fitting there will always be some errors,resulting in errors in the prediction results,after several simulation experiments and access to information found that there is a sudden burst of network traffic,the quality of the original data Large problems,which led to our in the process of fitting the model often white noise test does not pass,the parameter factor is not significant,the prediction results are not accurate and many other problems eventually lead to experimental failure.Therefore,finding the right algorithm to improve the quality of the data is the key to the correct fitting of the model.The main work of this paper is to introduce the theoretical basis and research contents of FARIMA model in detail in the first two chapters,and analyze the important characteristics of FARIMA model,such as self-similarity and long correlation.It is precisely because FARIMA model has the characteristics of long memory The prediction of network traffic data has laid a good foundation for traffic forecasting.Secondly,in order to solve the quality problem of network traffic data,this paper proposes two abnormal value processing algorithms in the third chapter,one is missing value padding algorithm,the other is the elimination algorithm of outliers and uses the core router of Hunan Radio and TV The experimental results show that the data quality verificat ion algorithm improves the accuracy of the data and reduces the data error,which makes the original data more complete and the fitting model is more accurate.In the fourth chapter,the FARIMA model is used to simulate the real data of Hunan Radio and Television with the FARIMA model.The results show that the FARIMA model with data quality test is more accurate and has stronger data prediction ability The Which laid a solid foundation for the future network traffic forecast.The last chapter summarizes the research contents and algorithms of the whole paper,points out the shortcomings of the current research contents,and prospects the future research.