190
Chapter 9
Todesignate a branch as the scoring branch (Tools menu)
EConnect the model nugget to a terminal node (a processing or output node downstream from
the nugget).
EOn the main menu, click:
Tools> Stream Properties > Deployment
EOn the Deployment type list, click Scoring Only or Model Refresh as required. For more information,
see the topic Stream Deployment Options on p. 185.
EClick the Scoring node field and select a terminal node from the list.
EClick OK.
Model Refresh
Model refresh is the process of rebuilding an existing model in a stream using newer data.
The stream itself does not change in the repository. For example, the algorithm type and
stream-specific settings remain the same, but the model is retrained on new data , and updated
if the new version of the model works better than the old one.
Only one model nugget in a stream can be set to refresh—this is known as the refresh model. If
you click the Model Refresh option on the Deployment tab of the stream properties (see Stream
Deployment Options on p. 185), th e model nugget that you designate at that time becomes the
refresh model. You can also designate a model as the refresh model from the pop-up menu of a
model nugget. The nugget must already be on the scoring bran ch for this to be possible.
If you turn off the “refresh model” status of a nugget, this is equivalent to setting the deployment
type of the stream to Scoring Only, and the Deployment tab of the stream properties dialog box is
updated accordingly. You can turn this status on and off by means of the Use as Refresh Model
option on the pop-up menu of the nugget on the current scoring branch.
Removing the model link of a nugget on the scoring branch also removes the “refresh mo del”
status of the nugget. You can undo removal of the model link by means of the Edit menu or the
toolbar; doing so also reinstates the “refresh model” status of the nugget.
How the Refresh Model is Selected
As well as the scoring branch, the link to the refresh model is also highlighted in the stream. The
model nugget chosen as the refresh model, and therefore the link that is highligh ted, depends on
how many nuggets are in the stream.