About the scikit-learn API

This is a question about regression using scikit-learn's decision tree.

In this algorithm,
.score and
predictive accuracy The .features_importances_indicates the importance of feature quantities is
I'm sure there are, but
How is each regression calculated?

For classification, the former is the percentage that was correctly classified. I remember that the latter is a characteristic amount that greatly reduces the degree of impurities...
How is it expressed in regression?

I would appreciate your encouragement.