Tools for graph mining

Large real-world graphs often show interesting properties, such as power-law degree distributions and very small diameters. Discovering such patterns and regularities has a wide range of potential applications. It could help us with detecting outliers or abnormal subnetworks (such as terrorist networks or illegal money-laundering rings), maximizing e ciency of disease controlling, marketing, forecasting and simulations, to name a few. A graph generating model, the recursive matrix (R-MAT) model is introduced and a method (AutoMAT) is shown for automatically estimating the input parameters for R-MAT in order for it to match a given realworld graph. Using these parameters, the resulting R-MAT graphs are shown to match many properties of real graphs. Also, a set of plots (A-plots) are introduced as original ways for viewing large graphs. Their applications in nding interesting patterns and outliers in real, large graphs are demonstrated.