Maximum Likelihood Methods for Data Mining in Datasets Represented by Graphs

Due to the boom in complex network research, large graph datasets appeared in various fields, from social sciences (P. Holme et al., 2004) to computer science (C.R. Myers, 2003), (M.Faloutsos et al., 1999), (A-L. Barabasi and R. Albert, 1999) and biology (L. Negyessy et al., 2006). There is an increasing demand for data mining methods that allow scientists to make sense of the datasets they encounter. In this paper, we present two graph models and two maximum likelihood algorithms that fit these models to pre-defined data. We also show two example applications to illustrate that these algorithms are able to extract interesting and meaningful properties from the data represented by appropriate graphs.