Is there a way to make partial dependence plots for random forest with multiple classification in Python (using scikit-learn)?

Question

Is there a way to make partial dependence plots for the random forest with multiple classification in Python (using scikit-learn)?

I'm raising a separate question about this because I'm not sure if such a function exists in scikit-learn. I've seen a few examples in R already. If the function doesn't exist, I will make the request in scikit-learn github, but just want to double-check with the community before making the request.

If you know of any other Python package other than scikit learn that could conduct the plot, please let me know. Thanks.

Answer 1

You have to specify the class for which you want to plot the partial dependencies. This is done by the parameter "target" in the plot_partial_dependence function

For example, if you have three target classes "low", "medium", "high", you would say plot_partial_dependence(estimator, ..., target='high').

However, I'm still trying to find some answers regarding the interpretations of partial dependency plots for multi-class-classifiers. If you have some information, let me know.