The network mining plug-in was developed in order to process attributed graphs such as social networks, co-authorship networks, etc. The plug-in provides the infrastructure to handle large networks within KNIME as well as KNIME nodes to create, generate, manipulate, analyze, and visualize these networks.
The network infrastructure was developed to support a wide range of network types such as
- multi graphs,
- hypergraphs, and
- k-partite graphs.
It also enables attributes to be assigned (e.g. edge weights, node labels, network descriptions) - which we refer to as features - to all network elements such as
- the graph itself,
- edges, and
- end objects (pair that groups a node and an edge).
- string e.g. txt such as a label
- boolean (true/false)
- float e.g. edge weight
- uri e.g. the uri of an integrated data source
- Lists of the above data types e.g. string list
- enum with predefined values
The plug-in supports also the interactive viewing of networks and the sending and receiving of networks to and from external programs. For a list of the supported programs see the Network Visualization section.
For a detailed description of the available example workflows that demonstrate some of the nodes and the usage of the network mining plug-in, see the Examples section.
The network mining plug-in is an offspring of the BisoNet plug-in which has been developed for the EU Project Bison. The major goal of the plug-in was to provide on the one hand a graph structure that is flexible enough to support a wide range of graph structures (e.g. (un)directed, (un)weighted, hyperedges etc.) and on the other hand an infrastructure for handling large networks. The BisoNet plug-in has been actively used throughout the three-year project.