Enhanced OMP for Missing Traffic Reconstruction based on Sparse SVD

Abstract

Missing large amount of internet data is a crucial issue to be addressed in network monitoring. The missing information should be restored using only a minimum knowledge of the data. Compressive Sampling (CS) algorithm provides a solution to complete data by utilizing the properties of randomness in the input data. Recently the reconstruction algorithm has developed in the base dictionary using orthogonal based operators. In this paper, we consider a CS approch to solve the missing problem using Singular Value Decomposition (SVD) sparsity, routing matrix for measurement matrix, and Orthogonal Matching Pursuit (OMP) as a recovery algorithm. To improve the accuracy, we also incorporating linear interpolation after OMP and Bilinear interpolation after SVD reconstruction. The missing scheme is randomized to simulate the actual behaviour of the network. Our experiments show that our proposed method is capable to fix large missing values with a high degree of accuracy for all missing type. This method is superior compared to the method in previous studies.