Finding associations-between-groups in multimode networks

A multimode network involves more than one type of entities. In a multimode network, besides nodes with the same type can form groups, nodes in different groups with the same type or the different types also tend to form bigger groups because nodes in these different groups frequently are related. These bigger groups (named as Associations-between-Groups, for short AGs) provide more information which can be used to reveal the properties and the behaviors about entities in different groups. This paper supposes each type of nodes have been grouped by using existing methods, and finds associations between groups with the same type or the different types in a multimode network. Firstly, the paper defines related concepts of AGs, including row instance, participation ratio and participation index. Then the paper proposes a general method to generate AGs and AGs´ row instances, and a basic algorithm (called APAG) to find prevalent AGs. The paper also presents an efficient algorithm (called APLink-AG) to find prevalent Link-AGs (special AGs) by using the anti-monotone of prevalent measures. Finally, experiments on synthetic data evaluate algorithms APAG and APLink-AG, and show APLink-AG is efficient.