A sparsity adaptive measurement algorithm for network traffic matrix

In order to improve the accuracy of the measurement algorithm for traffic matrix, a novel traffic matrix measurement algorithm with compressive sensing is proposed. This algorithm gets the judge gate by the principal components analysis and normalization of singular value. To reduce the measurement error created by approximation of sparse express and inaccurate choice of sparsity, we use L2 formulation of residual error to match the sparsity in the process of reconstitution of the traffic matrix on each time of measurement. Simulation results show that, this algorithm can obtain less spatial relative error and temporal relative error compared with the existing algorithm. With the help of adaptive selection for initial value of sparsity, this algorithm can obtain a higher accuracy.