If you build models in Pipeline Pilot on a regular basis, you simply must learn how to use the Shortcut Component.

Here's why: When you run a protocol that uses a learner component to build a model, a new model component gets created in your user tab. To make predictions with the model in the same protocol, you run the protocol once, and then drag in the model component and send some data through it. The model component doesn't exist until you run the protocol the first time. The component comes in as a shortcut, so if you then export the protocol and send it to a colleague, your colleague will get a "master component cannot be found" error when opening it.

The Shortcut Component allows you to make predictions with a model within the same protocol used to build it, without having to include the model component in the protocol. It works for both calculable properties (as output by native PP models) and other properties (as output by R models). For calculable properties only, you can instead use the PilotScript calculate() function, but calculate() does not allow you to override any of the model's default parameter settings. The Shortcut Component does.

The Shortcut Component acts like a shortcut to any saved component in Pipeline Pilot. Just enter the component's name as the value of the Component Name parameter. (Use a full path such as myname/LearnedProperties/mymodel if there is any chance of ambiguity.)

But the really powerful feature is that you can add parameters to the Shortcut Component in order to override the model's defaults. Just right-click the component and select "Edit...". Then add the parameters. The only tricky thing to watch out for is when you want to override the default Output parameter value for a model. Output is a "special" parameter in Pipeline Pilot, so you can't just specify its values like you would for a normal parameter. To override default model Output values, you have two options:

1. Run the protocol once to generate the model. (If your training set is large, just send a small fraction of it into the learner, since all you care about at this point is the output names.) Drag in the model. Right click the model and select "Edit...". Copy the Output parameter to the clipboard. Edit the Shortcut Component, and paste the Output parameter. Click OK and then change the selected outputs as you desire.

2. Edit the Shortcut Component. Create a parameter named Output1 of type String. Click the Array box. Enter the desired output properties in the Legal Values tab. Right before you click OK, delete the "1" from the parameter name. Click OK again to close the Edit dialog, and set the selected outputs to the ones you want.

I find the first method simpler, because then I don't need to figure out what the allowed output names will be (based on the model name and learner settings).

The attached PP 8.0 protocol shows an example of using the Shortcut Component as described here. In addition, all of the components in Pipeline Pilot 8.0 that make use of model applicability domain information (such as the Model Applicability Filter and Regression Model Evaluation Viewer) make use of the Shortcut Component. These components are subprotocols, so you can open them up and look inside if you want to see how they work.