Clustering analysis measures the overall connectedness of a graph and how much individual nodes cluster together. A cluster tends to be a group of highly connected nodes that share common neighbors.
To perform cluster analysis, complete the following steps:
Click Link Analysis on the Link Chart Diagram tab to open the analysis overlay window.
To minimize the window, click the arrow at the upper left.
- Under Analysis Method, choose Clustering.
- Under Analysis Type choose one of the following options:
- Biconnected Components—Partitions the network into clusters by analyzing its biconnected components. A biconnected component is a subgraph that is connected and nonseparable. This means that if a single node is removed, the component is still connected. Nodes are clustered so that nodes within each cluster are biconnected. Nodes belonging to multiple biconnected components are assigned to exactly one cluster.
- Edge Betweenness—Partitions the network into clusters using edge betweenness centrality. To determine this, iterations occur in which the edge with the highest betweenness centrality is removed from the network. Iterations cease if there are no more edges to remove or if the requested maximum number of clusters is found. The clustering with the best quality is returned.
- Hierarchical—Partitions the network into clusters using hierarchical clustering. Hierarchical clustering creates a hierarchy of clusters in a bottom-to-top approach based on a distance metric and linkage. The clustering is performed using the agglomerative strategy (for example, bottom-up) according to which clusters each node belongs. At each step, clusters are merged while moving up the hierarchy. The dissimilarity between clusters is determined based on the given linkage and given node distance metric.
- K-Means—Partitions the network into K clusters based on their position so that the node's distance to the clusters' mean centroid is minimized.
The analysis runs and the tabular results appear in the analysis overlay window.