A method to extract patterns of activity from time series of data structured as a network of objects, the method comprising the steps of creating a base graph of similar or related items from a given corpus by comparing intrinsic features of the items, combining time varying data on nodes or edges of the base graph, creating dynamic activity patterns by connecting nodes of the base graph according to a measure of activity from the time series of data, and creating static activity patterns by folding the dynamic activity patterns according to identifiers of the nodes from the base graph.