You are viewing the RapidMiner Studio documentation for version 10.0 - Check here for latest version
Generate Power Transform
(Operator Toolbox)
Synopsis
This operator applies (invert) BoxCox or (invert) Yeo-Johnson transformation on the selected attributes.Description
In some cases it can be benefitial, if a given column follows a normal distribution rather than any other distribution. Power Transformations are designed in a way, to turn a given distrtribution into a distribution which is similar to a normal distribution.
This operator implements two methods of power transformation. The standard and most known method is the BoxCox method. This method is very well known, but does not work on negative values.
The other method implemented is Yeo-Johnson, which also works for negative values.
For more information about the transformations as well as their respective definitions, please see: https://en.wikipedia.org/wiki/Power_transform
Input
exa (Data Table)The input ExampleSet which contains the data to be transformed
Output
- out (Data Table)
The ExampleSet with transformed attributes
- ori (Data Table)
The original ExampleSet.
mod (Preprocessing model)A model to apply the same transformation on a different data set.
Parameters
- method Defines the method. Uses BoxCox transformation. Only works for positive values. Uses Yeo-Johnson, which is the extension of box-cox for negative values. Inverse function of the BoxCox transformation. Inverse function of the Yeo-Johnson transformation. Range:
Tutorial Processes
Using Box Cox with ARIMA
In this example we take a shortend version of the German gas station data set. We first apply a box cox transformation and then fit and forecast the time series with ARIMA. Later we transform the forecast back into the normal scale using an inverse box cox transformation
