Kohonen Model Nuggets

Kohonen model nuggets contain all of the information captured by the trained Kohenen network, as well as information about the network's architecture.

When you run a stream containing a Kohonen model nugget, the node adds two new fields containing the X and Y coordinates of the unit in the Kohonen output grid that responded most strongly to that record. The new field names are derived from the model name, prefixed by $KX- and $KY-. For example, if your model is named Kohonen, the new fields would be named $KX-Kohonen and $KY-Kohonen.

To get a better sense of what the Kohonen net has encoded, click the Model tab on the model nugget browser. This displays the Cluster Viewer, providing a graphical representation of clusters, fields, and importance levels. See the topic Cluster Viewer - Model Tab for more information.

If you prefer to visualize the clusters as a grid, you can view the result of the Kohonen net by plotting the $KX- and $KY- fields using a Plot node. (You should select X-Agitation and Y-Agitation in the Plot node to prevent each unit's records from all being plotted on top of each other.) In the plot, you can also overlay a symbolic field to investigate how the Kohonen net has clustered the data.

Another powerful technique for gaining insight into the Kohonen network is to use rule induction to discover the characteristics that distinguish the clusters found by the network.

For general information on using the model browser, see Browsing model nuggets