Multithreading in `apply_forrest`

[pfarndt]

I just did some test and realized that when using many threads, e.g. 32, and handling RandomForestRegressors with many trees, e.g. 1000, it is way faster to build a forest than to apply a forest to get a prediction.

I think this is due to the fact that multiple threads are used to build the forrest in parallel, however the "application" of a forest is done sequentially. Would it be conceivable to also apply a forest using multiple threads?

This this issue could also be related to this issue

[rafaqz]

Not having this makes julia look really bad in comparison to Python. Is there any reason this can't happen?

[ablaom]

Multithreading in prediction was added in https://github.com/JuliaAI/DecisionTree.jl/pull/188/files . You need to buy in by specifying use_multithreading=false in apply_forest(...).

Closing, but feel free to open a new issue if you are still seeing issues.

[rafaqz]

I made a PR documenting this if you can merge it?

Without docs the feature doesn't really exist.