1

I'm currently learning the MLR package. MLR provide function to enhance the power of base learner. Like makePreprocWrapperCaret for data preprocessing, makeFilterWrapperfor feature selection. I'm wondering if there are any more flexible functions, that can allow the merge of complex processing in one learner, like first scaling the data, then feature selection, and finally tune the data. Or I have to write my custom learner function to achieve this?

Jiahao-AI
  • 39
  • 4
  • 1
    Have a look at [mlrCPO](https://github.com/mlr-org/mlrCPO) for setting up advanced ML pipelines with mlr. – pat-s Jan 10 '20 at 08:55
  • In case you are not aware of it, `mlr` has been retired in favor of `mlr3`. The latter (in combination with `mlr3pipelines` comes with more powerful features for setting up complex ML workflows. – pat-s Jan 10 '20 at 08:56
  • I knew that. But it pretty hard managing those S6 objects. I'll see if I can build a workflow that fits my needs in MLR first, which I'm more accustomed to. – Jiahao-AI Jan 10 '20 at 16:32
  • Its R6, not S6. Then mlrCPO is worth a look ;) – pat-s Jan 10 '20 at 18:07

1 Answers1

0

You can easily do this by layering multiple wrappers. For more information on wrappers, see the relevant part of the tutorial.

Lars Kotthoff
  • 107,425
  • 16
  • 204
  • 204