Graph Clustering For Viewer

Computes a hierarchical clustering of a logical graph containing 2 levels. The hierarchy levels are written as separate logical graphs to the given path, each to a folder starting with 'level_', e.g. 'level_0' for the lowest level. The top level should have significantly fewer elements than the original graph, which helps to visually explore very big graphs with the Big Graph Viewer node.

The superordinate hierarchy level consists of super vertices, where each is a parent for a number of vertices from the next subordinate hierarchy level and holds references to its children. Vertices in a subordinate level are first assigned to a cluster. Vertices in a cluster are than grouped by label to preserve both, the structure and the semantics of the graph. Each group then forms a super vertex. All edges between super vertices are combined for each direction.

How well the top level elements are reduced depends on the original graph structure and the type of cluster assignment algorithm.


Selected directory / Sink path
Directory to write the hierarchy levels
Overwrite existing files
Overwrite existing files?
Data sink type
The format to save the hierarchy level graphs
Cluster assignment algorithm

How the vertices are assigned to a cluster / community:

a) Label Propagation: the Flink Gelly Label Propagation algorithm

b) Structure Based Propagation: a custom Label Propagation algorithm, which uses structural properties. Computational more demanding

c) Group By Label: groups all vertices by label for the whole graph. Builds small top level but destroys local graph structures like communities.

Max iteration
the max number of iterations for the cluster assignment

Input Ports

A single logical graph.

Output Ports

Synthetic output, needs an execution node to run the Gradoop job.

Popular Predecessors

Popular Successors


This node has no views


  • No workflows found



You want to see the source code for this node? Click the following button and we’ll use our super-powers to find it for you.