CategoricalEncoding Class

[SerializableAttribute]
public abstract class CategoricalEncoding

<SerializableAttribute>
Public MustInherit Class CategoricalEncoding

[SerializableAttribute]
public ref class CategoricalEncoding abstract

[<AbstractClassAttribute>]
[<SerializableAttribute>]
type CategoricalEncoding = class end

Inheritance

Object  →  CategoricalEncoding

Use the CategoricalEncoding class to specify how a categorical variable should be encoded as indicator variables. An encoding is a set of mappings from categories to real numbers. The mappings are returned as Index<T> objects.

Encodings can have two forms: full rank and reduced rank. A full rank encoding means that, when the encoding is applied to each category, the resulting matrix has full rank. Reduced rank encoding means that the resulting matrix does not have full rank. This is useful when specifying models that include a constant term. In such cases, the full rank encoding and the constant term are linearly dependent.

To create an encoding, use one of the static methods of the CategoricalEncoding class. You can then use the GetFullRankEncoding(Int32) or GetReducedRankEncoding(Int32) to get the encoding. Alternatively, you can inherit from CategoricalEncoding

The creation of indicator variables is mostly transparent. To create an indicator variable, use the index with the WithCategories<U>(IList<U>) method.

Reference