Extracting baseline patterns in Internet traffic using Robust Principal Components

Baselining network-wide traffic by time-frequency constrained Stable Principal Component Pursuit

“…For simplicity, we assume each time-series N j (1 ≤ j ≤ P) is the white Gaussian noise with variance σ 2 j > 0 in this study 3 .…”

“…We verified that all these matrices satisfy the hypotheses of the R-TMDM model, the rank of each baseline matrix is 11, and its critical frequency f c = 112 2016 . In this study, SPCP-TFC is compared with other two network-wide traffic baseline schemes: RBL [3] and PCA [1]. For the PCA scheme, the number of principal components is fixed at 11, which is equal to the rank of the ground-truth baseline matrix.…”

“…Usually, the total traffic is modeled by the superposition of diverse components corresponding to different user behaviors [1], [2], [3], [4]. The baseline traffic represents the most prominent traffic patterns [3], which is quite helpful for many significant network applications such as capacity planning, load balancing, and anomaly detection.…”

“…Usually, the total traffic is modeled by the superposition of diverse components corresponding to different user behaviors [1], [2], [3], [4]. The baseline traffic represents the most prominent traffic patterns [3], which is quite helpful for many significant network applications such as capacity planning, load balancing, and anomaly detection. In the past, most studies devoted to estimating the trend of the single-link traffic [5], rather than extracting the common traffic pattern of the whole network, however, the later is more informative for the manager of a large-scale network.…”

Identification of tidal-traffic patterns in metro-area mobile networks via Matrix Factorization based model

“…A number of researchers are looking particularly closely at the application of Machine Learning techniques (a subset of the wider Artificial Intelligence discipline) in order to identify trends that are frequent in the network. Methods such as K-means, Spectral Clustering, Principal Component Analysis [4] and Gaussian mixture model have been exploited to identify and extract the basic traffic pattern and capture the underlying traffic trend. A range of applications, as anomaly detection and load balancing, rely on basic pattern estimation, and it represents an important solution for those Internet Service Providers (ISPs) that face every day network congestion problems or over-provisioning of the link capacity.…”