Research On Methods Of Network Measurement Data Processing Based On Tensor Completion

In recent years,with the rapid development of personal devices,accurate network measurement plays a key role in the field of networking research.For example,in network management,network security and the other fields,complete network latency data can analyze the network data transmission,so as to effectively deal with network failures and achieve better network performance optimization.However,in the actual network measurement system,data missing and data anomaly are inevitable.These problems not only seriously affect the accuracy and reliability of networkmeasurement,but also lead to making wrong research conclusions.Therefore,how to recover the original data from the incomplete network measurement data has become a hot topic for researchers.Among many network measurement,network latency is one of the research objectives that many researchers pay most attention to.In this paper,the processing method of typical network measurement data is studied,and a tensor completion method for network measurement is proposed.The main contributions are summarized as follows:(1)Aiming atrecovering network latency from incomplete data of personal devices,this paper constructs a tensor completion framework based on t-SVD,and takes into account the potential spatial information in the network latency data,and introduces the Laplacican regularization constraint,which effectively improves the recovery accuracy of latency data.Finally,a large number of simulations were performed using the real Seattle data set,and the results demonstrate the effectiveness of the proposed algorithm.Compared with the existing methods,the performance of the proposed algorithm achieves better performance than other algorithms in different structural missing patterns.(2)Aiming at dealing with missing and abnormal network latency data of personal devices,this paper constructs a tensor robust principal component analysis model based on t-SVD.The tensor robust principal component analysis is carried out on the network latency tensor data with outliers,which is decomposed into low-rank approximate tensor and abnormal sparse tensor.At the same time,the spatial characteristics of network latency data are analyzed,and the Laplacican regularization constraints are constructed,which can effectively improve the recovery accuracy and anomaly detection of network latency data.A large number of simulations are carried out using Seattle data set,and the results show that the recovery performance and anomaly detection performance of the proposed algorithm in terms of recovery accuracy and anomaly detection rate outperforms the other algorithms.