2

I'm learning with RapidMiner and I would love to run Linear Regresion on my data set. But I'm getting error that this method/technique cannot handle polynomial attributes.

What is the best way how to figure out this problem? Those attributes are basic strings and very important inputs as well.

Is there some kind of technique how to convert those strings, polynomial attributes, into numbers or integers or whatever?

Petr Bečka
  • 774
  • 3
  • 16
  • 39
  • Can you let us see a sample of your data as well as some indication of the function that you want to fit? – Bill Bell Oct 20 '17 at 03:41
  • You may consider how to transform the polynomial data, multiply fixed terms may be a good start. – shane Oct 20 '17 at 08:19

1 Answers1

2

RapidMiner offers several Operators that help you to change the type of an Attribute.

You can find them under: Blending -> Attributes -> Types

In your case the best one choice would probably be the Nominal to Numerical. Typically you choose one of those two transformations, or coding types:

  • unique integers, which assigns each String value an unique number
  • dummy encoding, which creates a new Attribute for each value. This new Attribute is then either 0 or 1.

For a linear regression, I recommend using the dummy encoding, as it does not introduce any bias, based on the assigned number.

Also feel free to ask further, or re-post, questions in the RapidMiner community forum.

David
  • 792
  • 5
  • 17