What is the difference between R-squared generated during machine learning model training and the R-squared generated for model performance under Explanations?

kbennett 1

For some of my ML experiments, I've selected R-squared as the primary metric for guiding the training. I also noticed that under the Explanations section of the algorithm overview, there is a section called Model Performance and with it, a box plot graph that also shows a different value for R-squared.

What is the difference between these two values of R-squared?

romungi-MSFT 41,861 Reputation points Microsoft Employee

2022-11-23T04:29:01.683+00:00

@kbennett Does the selected model contain explanations for raw or engineered data features? I think this could be different explanation ids for raw or engineering data features and the one on the left is a metric from overall run.
kbennett 1 Reputation point

2022-12-02T18:13:31.05+00:00

Oh, ok. Yes. The model contains explanations for both raw and engineered features.

I read the whitepapers on explanations and their visualizations but it's still somewhat murky. Do you have a resource or are you able to explain what is meant by raw vs engineered features?