Asked & Answered: How can I override the champion model in a Model Studio machine learning project?

[#1]

Question

In Model Studio, I have created a Data Mining and Machine Learning project with multiple pipelines. The overall champion model listed on the "Pipeline Comparison" tab is a model that I suspect is largely overfit, so I do not want to use that model. How can I set another model to be the champion for my project?

Answer

Model Studio automatically determines the best model from each pipeline (that is, a pipeline champion) based on the selection criterion set in the Model Comparison node properties (note that the default for this can be changed in the Project Settings). Then, the Pipeline Comparison tab automatically lists all pipeline champions, compares them, and select an overall project champion model, as shown here:

Comparison of models across all pipelines

You have complete control to override the selected champion model for the project. To do this, simply select a model in the table and choose "Set as champion" in the menu in the upper right just above the table (also available by right-clicking on the model in the table).

In the case where the model you want to choose is not the champion for a pipeline (and thus is not listed in the Pipeline Comparison table by default), go back to the pipeline that created that model, click on the menu button (3 dots) on that node (or right-click the node) and select "Add challenger model" as shown here:

This will then present this model to you in the Pipeline Comparison table and allow you to select it as the champion model for the project as described above.