Traffic Matrix Completion by Weighted Tensor Nuclear Norm Minimization

Traffic Matrix (TM) plays an essential role in many network analysis task. Since obtaining whole TM by direct observation is challenging, a lot of studies on recovering TM from the partial observation. These methods achieve high recovery accuracy by using spatio-temporal characteristic of TM which also need to be estimated from the partial observation, and the approach leads to high computational cost and instability. We proposed a new TM completion method using a weighted tensor nuclear norm minimization with tensor construction based on the intrinsic periodicity of TM. Our tensor construction method does not require unstable spatio-temporal characteristic estimation from the partial observations of TM. The experimental results on real-world traffic data show that the proposed method can achieve the comparable recovery capability as the conventional methods with a significantly simple problem formulation.