How to best do cross-validation using fitensemble?

Question

Tobias Pahlberg 2016-6-23

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/291816-how-to-best-do-cross-validation-using-fitensemble

回答： Don Mathis 2017-3-31

Hi

I generated code from the Classification Learner app where I wanted to cross-validate a classifier. It gave me something like:

 classificationEnsemble = fitensemble(predictors, response, Bag', nRounds, Tree', type', 'Classification', ...); 
 partitionedModel = crossval(classificationEnsemble, 'KFold', 5);

But I can also provide the cross-validation parameter directly into the fitensemble command. Like:

classificationEnsemble = fitensemble(..., 'KFold', 5);

Is there any difference here? Is the first case trained on all the data and then cross-validated??

/Thanks

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Don Mathis 2017-3-31

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/291816-how-to-best-do-cross-validation-using-fitensemble#answer_261229

It doesn't make any difference in the models. The only difference is that with the first method you also get a single model trained on the full dataset.