Article Preview
TopI. Introduction
Real life networks are increasingly being modeled as graphs. Graphs provide an excellent representation of interconnections amongst the nodes of a network. Apart from the information depicted by these connections, some information lays unapparent amongst its properties. Thus, it is imperative to develop methods to process graphs to mine such information. A large amount of work has been done in the field of mining this data from such interconnections.
These connections have traditionally been modeled as static networks (Elseidy et al., 2014; Guimei et al., 2009; Inokuchi et al., 2000; Inokuchi et al., 2003; Kuramochi & Karypis, 2001; Kuramochi & Karypis, 2004; Yan & Han, 2002), which do not change over time. However, real life networks are dynamic in nature, where the nodes and their connections evolve with time. Thus, nowadays, networks are increasingly being modeled in a time-series representation (Borgwardt et al., 2006; Desikan & Srivastava, 2004; Duan et al., 2009; Gupta & Thakur, 2013; Gupta, Thakur, & Goel, 2014; Gupta & Thakur, 2013; Gupta & Thakur, 2015; Gupta, Thakur, & Gundherva, 2014; Gupta et al., in press; Gupta, Thakur, & Kishore, 2014; Halder et al., 2013; Holder & Cook, 2009; Lahiri & Berger-Wolf, 2010; Lin et al., 2008; Obulesu et al., 2014; Rasheed et al., 2011; Yang et al., 2014) where it is represented as a series of graphs. Each subgraph is a snapshot of the network at successive intervals within the duration of observation (Yang et al., 2014). For such representations, data in the graph is seen as data sequence of occurrence, indegree, outdegree etc. for each node and edge. The length of the sequence is equal to the number of snapshots, and each element in the sequence corresponds to the status of the node, or edge, for that snapshot.
Such dynamic graphs usually exhibit recurring patterns in their data-sequences. These recurring patterns find highly lucrative possibilities in domains of email, social networks, transportation networks, stock markets, and many more. They can be used to predict information such as share-price trends, road rush hours, vacation hotspots, airline traffic etc. Extensive research has been carried out for discovery of such patterns(Borgwardt et al., 2006; Gupta & Thakur, 2013; Gupta, Thakur, & Goel, 2014; Gupta & Thakur, 2013; Gupta & Thakur, 2015; Gupta, Thakur, & Gundherva, 2014; Gupta et al., in press; Gupta, Thakur, & Kishore, 2014; Halder et al., 2013; Holder & Cook, 2009; Lahiri & Berger-Wolf, 2010; Obulesu et al., 2014; Rasheed et al., 2011) within evolving network domains. But much attention has not been given to the behavioral analysis of such patterns, i.e. the application and interpretation of the patterns for the said domains.