How to get model performance on the train set ?

[MatthieuPx]

Within a lab/visual analyses/classification task, after a model was trained, how to get the model performance on the train set ? Is there a direct way to get it without deploying the model and evaluating it on the train set ?  

I believe that all displayed metrics in the Results tab are (obviously) related to the test set. What if we need to look at some metrics (AUC or else) on the train set, to check at the overfitting ? 

Thank you. 

[AdamN · Accepted solution]

Hi @MatthieuPx, thank you for your question.

You will indeed need to deploy the model to do this.

You can use an Evaluate recipe to compute the model's performance on the train set. In the settings for the Evaluate recipe, you can choose which metrics to compute.

In case you do not have the train and test sets in separate datasets, you can export the train and test sets from the model results, as shown here:

Also, we are currently working on adding learning curve charts which will show the model's performance on the train and test set. This will be available in a future version of Dataiku.