Network Traffic Prediction Based On Gray Prediction And Markov Process

Network traffic prediction and anomaly detection have been researched excellent performance,but these researches are mainly based on large data analysis which has higher demand in resources and computing ability and storage capacity of network devices.Therefore,the existing prediction and anomaly detection technology may be not applicable to the energy,resource constrained network equipment.However,with the popularity of Internet of Things,it is necessary to study the network traffic forecast and anomaly detection technology which can be applied to resources constrained equipment to ensure the rational allocation of network resources and the security of the network.Gray prediction model has the advantages of small sample,poor information and low computational complexity.Small sample and poor information indicate that the model has low demand for data and requires very limited storage resources.Low computational complexity means that the required computing resources and energy are limited.Therefore,the algorithm's spatial and temporal complexity is relatively low.This paper mainly discusses the gray prediction theory used in network traffic prediction and anomaly detection,especially for energy,resource-limited Internet of things.This article mainly includes the following aspects:Firstly,the thesis introduces the basic gray forecasting model,including the establishment conditions of the model,the accuracy of the model and the prediction.Secondly,the characteristics of the network traffic are studied,and the feasibility of gray model in traffic modeling is verified.On the basis of the gray prediction model,the data preprocessing including different buffering operators on the model prediction is analyzed.And the buffer operator with optimal prediction effect is determined.Experiments show that the exponential weakening buffering can improve the accuracy of prediction based on the basic gray prediction for the network traffic.Thirdly,the basic gray prediction model possesses higher prediction accuracy,but cannot achieve continuous updates.In this paper,the long-term continuous prediction is realized by the metabolic gray prediction model.Combined with the previous data preprocessing,the metabolic gray prediction model based on the exponential weakening buffer operator?ewboMGM1,1?was determined.Experimental analysis of the model was made using the UK data set.The results show that the model has higher prediction accuracy and faster prediction speed.At the same time,the trend prediction of network traffic based on Markov process is presented to predict the trend of future traffic in the form of probability,which makes up the shortage of ewboMGM_1,1.Finally,based on the proposed ewboMGM_1,1 model,the method of detecting abnormal network traffic based on multidimensional statistical features is studied.The KDD data set is used to compare the proposed method with other two commonly used anomaly detection methods.The experimental results show that the proposed method achieves the compromise between time and accuracy,and reduces the anomaly detection time on the basis of ensuring the accuracy rate.