2018/05/16 by Yannick Léo, Léo, Yannick, Christophe Crespelle +3
Physics and Astronomy · Computer Science · #Complex Network Analysis Techniques #Data Visualization and Analytics #Data Management and Algorithms
paper · pdf · doi:10.48550/arxiv.1805.06188
Many dynamic networks coming from real-world contexts are link streams, i.e.\na finite collection of triplets (u,v,t) where u and v are two nodes\nhaving a link between them at time t. A very large number of studies on these\nobjects start by aggregating the data in disjoint time windows of length\n\Δ in order to obtain a series of graphs on which are made all subsequent\nanalyses. Here we are concerned with the impact of the chosen \Δ on the\nobtained graph series. We address the fundamental question of knowing whether a\nseries of graphs formed using a given \Δ faithfully describes the\noriginal link stream. We answer the question by showing that such dynamic\nnetworks exhibit a threshold for \Δ, which we call the \saturation\nscale, beyond which the properties of propagation of the link stream are\naltered, while they are mostly preserved before. We design an automatic method\nto determine the saturation scale of any link stream, which we apply and\nvalidate on several real-world datasets.\n