Compute partial dependencies from learned trees

[ChristophAymannsQC]

Summary

Partial dependencies can be computed directly from the data distribution in the leaves of the trees in the booster, rather than by calling predict, see references below. This tends to be faster. It would be nice if LightGBM allowed for direct computation of the partial dependencies using this method.

Motivation

Partial dependency plots are widely used to interpret a model. Providing a fast method for computing them would be a big asset.

References

• A good overview of the method can be found here: http://nicolas-hug.com/blog/pdps.
• The original idea goes back to Friedman, Jerome H. 'Greedy function approximation: a gradient boosting machine.' Annals of statistics (2001): 1189-1232.

[StrikerRUS]

Closed in favor of being in #2302. We decided to keep all feature requests in one place.

Welcome to contribute this feature! Please re-open this issue (or post a comment if you are not a topic starter) if you are actively working on implementing this feature.

[jameslamb]

Sorry, this was locked accidentally. Just unlocked it. We'd still love help with this feature!