How to use Bayesian Optimization?

I'm trying to run the following Mathworks example with my own X and Y:

"Tune Random Forest Using Quantile Error and Bayesian Optimization"

https://jp.mathworks.com/help/stats/tune-random-forest-using-quantile-error-and-bayesian-optimization.html

But, I'm getting the following error:

Undefined function or variable 'Y'.

I have attached the modified code (place both files in one folder on your PC drive). Can anyone help?

Answer 1

[Don Mathis]

You need to pass Y into oobErrRF. Change its first line to

function oobErr = oobErrRF(params,X,Y)

And change the call on line 66 of your main file to

results = bayesopt(@(params)oobErrRF_editted(params,X,Y),hyperparametersRF,...

That fixes your error. But after that you get a new error, because inside oobErrRF you're calling oobQuantileError on a classification random forest, while it's only defined for regression random forests. Are you trying to do classification or regression?

[Amy Xu]

Thanks @Don.

In fact, I am trying to do classification. Can we use Bayesian Optimization for classification?

[Don Mathis]

Yes you can. I edited your code to call 'oobError' instead of 'oobQuantileError', and took the mean over all trees. I also told your final 'Mdl' to train with Method 'classification' and turned on 'OOBPredictions' so you can see the performance of the final model. I also told 'bayesopt' to use Verbose=1. I've attached the edited files.