Mining Closed Relational Graphs with Connectivity Constraints

Relational graphs are widely used in modeling large scale networks such as biological networks and social networks. In a relational graph, each node represents a distinct entity while each edge represents a relationship between entities. Various algorithms were developed to discover interesting patterns from a single relational graph (Z. Wu et al., 1993). However, little attention has been paid to the patterns that are hidden in multiple relational graphs. One interesting pattern in relational graphs is frequent highly connected subgraph which can identify recurrent groups and clusters. In social networks, this kind of pattern corresponds to communities where people are strongly associated. For example, if several researchers co-author some papers, attend the same conferences, and refer their works from each other, it strongly indicates that they are studying the same research theme.